Skip to content
Corporate & Investment Banking My progress Français

Home / Capital Markets & Securities Services / Capital Markets / Module 12

12Commodities Trading

From Price Discovery to Physical Settlement · Energy, Metals and Agriculture · The Commodity Trade Lifecycle, Paper and Wet

Energy, metals and crops, traded on paper and delivered for real.

Reading time
2 h 30
Sections
51
Recall questions
54
Quiz questions
20
Simulators
1

This module teaches commodities from zero, but goes deep enough for a professional

It assumes no prior knowledge of trading floors and builds every idea from its foundations. What a commodity is. Why anyone would buy oil they never intend to receive. Why a barrel loaded in West Africa and a barrel quoted on a screen in London are the same product at two different prices. Why a tonne of copper is a document, why electricity cannot be stored at any price, and why a wheat contract answers to weather and governments before it answers to markets.

Wherever a concept is technical, we stop and explain it in plain language before using it again. Terms are defined once, precisely, and then used freely. By the end you will read a trade ticket, a bill of lading and a margin call and understand what each one commits the bank to.

Two threads run through everything: the business and the technology

The business thread covers what commodity trading is, why producers and consumers need a bank in the middle, how a price is built from a benchmark rather than quoted from thin air, how physical cargoes move across oceans under contracts written decades ago, and how a desk makes money on the difference between two prices rather than on the direction of one.

The technology thread covers the systems that hold a commodity trade together. It follows a single deal ticket through the trade capture system, into risk and position management, out into scheduling and logistics, through confirmation, invoicing and settlement, and finally into regulatory reporting. Commodities is the one asset class where the IT platform must model a physical object that rusts, evaporates, spoils and sits on a ship for six weeks.

These two threads are inseparable. In commodities the process and the platform are the same thing: a trade that cannot be booked correctly cannot be hedged, and a hedge that the system cannot see does not exist.

How to read this module

Read it as a narrative, not a reference card

Each part assumes you have read the ones before it. Part 1 builds the market. Part 2 walks a single trade from first contact to final payment and is the spine of the module. Parts 3 and 4 take that same trade apart by actor and by system. Part 5 asks what goes wrong and what is changing.

If you are short on time, read the section headings on their own. Each one states a claim rather than naming a topic, so the argument survives being skimmed.

The module teaches five families and works one of them all the way through

Commodities is not one market. Energy, base metals, precious metals, gas and LNG, and agriculture each run on machinery of their own, and Part 1 sets out where each departs from the others: the London Metal Exchange's daily prompt dates and its warrants, electricity's sequence of day-ahead, intraday and balancing markets, the seasonality and destination optionality of gas and LNG, and the harvest cycle and political exposure of grains and softs.

From Part 2 onward the module follows a single physical crude oil cargo from first contact to final payment. That is a deliberate teaching choice rather than a narrowing of scope. The crude cargo is the only product that exercises every stage, every team and every system at once, so it is the most economical way to learn the machine. Once you can follow it, a copper cathode trade, an LNG cargo, a power hedge or a grain shipment differ in vocabulary far more than in mechanism, and Part 1 has already told you where the genuine differences lie.

Each coloured box plays a fixed role

Knowing what each box signals lets you skim or dive as you choose.

Key insight distils the one idea from a section worth remembering even if you forget everything else around it. Going deeper adds mechanism for readers who want the machinery rather than the summary. IT and data lens names the systems, data objects and interfaces that carry the step being described. Watch out flags a trap that has cost real institutions real money. Worked example carries the running case study, and every figure inside it is illustrative. In a nutshell closes each part with the argument compressed. Check your understanding gives you questions with answers, so you can test yourself rather than re-read.

What you will master

By the end of this module you will be able to do five things. Each maps to one part of what a commodities professional is expected to know.

You will understand the commodity trade lifecycle end to end

You will follow a trade from the market view that prompts it, through execution and trade capture, into hedging, chartering, loading, inspection, price fixing, invoicing and settlement, and out the other side into position management and regulatory reporting. You will know what happens at each stage, how long it takes, and which of those stages can never be undone. You will also know how that lifecycle changes when the commodity is a metal held as a warrant, an electron that cannot be stored, or a harvest that arrives once a year.

You will be able to map every actor

You will know who does what inside the bank and outside it. Producers, refiners, utilities, trading houses, hedge funds, exchanges, clearing houses, price reporting agencies, inspectors, shipowners and warehouse operators each play a defined role, and the desk sits in the middle of nearly every one of those relationships.

You will understand where the price actually comes from

You will know that a cargo price is almost never a number agreed in advance. It is a formula: a published benchmark, plus a differential for the grade, adjusted for quality, plus freight to the delivery point, averaged over a pricing period that may fall weeks after the deal was struck. You will be able to build that price yourself and explain which part of it the desk is actually taking a view on.

You will be able to trace data front to back

You will follow one deal ticket through the trade capture system, the risk engine, the scheduling platform, the confirmation stack, the invoicing system and the regulatory report, and you will see why a single mistyped pricing formula reaches all six.

You will recognise the risk and transformation landscape

You will know the six risk families a commodity desk carries, the regulations that shape every system in the chain, the three ways commodity trading has historically blown up, and which of today's technology changes genuinely move cost or risk rather than merely the interface.

1Context and the Big Picture

Before any process makes sense, you need the map. This part defines what a commodity is, places the trading desk inside the bank, lays out every product the market trades, and explains the cycle that decides whether the business is printing money or shedding headcount.

A commodity is a physical good so standardised that nobody cares who made it

The defining property is interchangeability, not usefulness

A commodity is a raw or primary good where one unit is treated as equivalent to any other unit of the same specification. A barrel of crude oil meeting a given assay is worth the same whoever pumped it. A tonne of copper meeting London Metal Exchange grade A specification is worth the same whoever smelted it. This property has a name: fungibility.

Fungibility is what makes trading possible at scale. If every barrel had to be inspected as a unique object, there could be no exchange, no benchmark and no forward market. Standardisation converts a physical object into a tradeable contract. That conversion is the foundation of the entire industry.

But the standardisation is never perfect, and the gap is where the money is. Crude oil is not one substance. It varies in density, measured as API gravity, and in sulphur content, which is what "sweet" and "sour" mean. A light sweet crude yields more valuable products per barrel and costs less to refine than a heavy sour one. The two are both crude oil, and they are not the same price.

Three families cover the entire field, and they behave differently

Energy covers crude oil, refined products such as gasoline, diesel and jet fuel, natural gas, liquefied natural gas, electricity, coal and carbon allowances. It is the largest by traded value and the most financialised. It is also where this module concentrates, because energy carries every complication the other families have and several of its own.

Metals splits in two. Base metals — copper, aluminium, zinc, nickel, lead, tin — are industrial inputs traded largely on the London Metal Exchange, which is unusual in that it still runs a genuinely physical delivery system through a global warehouse network. Precious metals — gold, silver, platinum, palladium — behave partly as industrial inputs and partly as monetary assets, which gives them a demand curve unlike anything else on this list.

Agriculture, often called softs and grains, covers wheat, corn, soybeans, sugar, coffee, cocoa and cotton. It is the family most exposed to weather, harvest cycles and government intervention, and the one where the same physical unit genuinely degrades over time.

Family Main products Principal venues What makes it distinctive
Energy Crude oil, refined products, natural gas, LNG, power, coal, carbon ICE, CME, plus a very large OTC and physical market Largest by traded value, most financialised, and the only family where one product (power) is effectively unstorable
Base metals Copper, aluminium, zinc, nickel, lead, tin London Metal Exchange, COMEX Genuine physical delivery through an approved global warehouse network, with title held as warrants
Precious metals Gold, silver, platinum, palladium COMEX, London OTC market Behave partly as industrial inputs and partly as monetary assets, giving them an unusual demand curve
Agriculture Wheat, corn, soybeans, sugar, coffee, cocoa, cotton CBOT, ICE Driven by weather and harvest cycles, heavily exposed to government intervention, and genuinely perishable

Three properties separate commodities from every other asset class

A commodity has to be somewhere. A share of stock exists in a register; it has no location. A cargo of crude is at a loading terminal, on a vessel, in a storage tank or in a refinery, and moving it costs real money and real time. Location is part of the price. This is why the same grade of oil has different values in Rotterdam and Singapore, and why a large part of the industry exists purely to move barrels from where they are cheap to where they are dear.

A commodity has to be stored, and storage is not free. Holding a barrel costs tank rent, insurance and financing on the capital tied up. Together these are the cost of carry. Holding electricity is close to impossible, which is why power prices behave more violently than anything else traded. Storability determines how far forward a market can price.

A commodity has a quality, and quality is measurable and disputable. Two cargoes of the same nominal grade can differ enough to matter, which is why independent inspectors exist, why contracts specify tolerances, and why a shipment can be rejected or repriced after it has already been loaded. No equity trade has ever been repriced because the shares arrived off-spec.

Each family runs on machinery of its own

The crude cargo teaches the pattern, and the other markets vary it in instructive ways

Part 2 follows a crude oil cargo because it exercises every stage of the lifecycle. A professional needs to know where the other families depart from that pattern, and the departures are not cosmetic. Each one changes the systems, the risks and the shape of the trade.

Base metals price to a date rather than a month, and title lives in a warrant

The London Metal Exchange is structurally unlike every other derivatives venue, because it grew out of a physical merchant market rather than a financial one. Most futures markets trade standardised monthly contracts. The LME trades daily prompt dates out to three months, then weekly, then monthly, a structure built so an industrial user can hedge to the exact day its metal is needed. The three-month date is the market's liquidity benchmark and it rolls forward every day, which means a position held without action does not stay where it was put.

Title works differently too. Metal held in an LME-approved warehouse is represented by a warrant, a document of title to a specific lot. A warrant can be traded, pledged as collateral, or cancelled for physical withdrawal. This makes metals financing structurally easier than oil financing, because the collateral sits still and has a document attached to it.

It also creates the market's characteristic pathology. When large volumes are cancelled for load-out and the warehouse's physical delivery capacity is limited, waiting times to withdraw metal can extend for months. A consumer who needs metal now must then pay a physical premium over the exchange price that reflects the queue rather than the metal. Warehousing practices and load-out rules have been the subject of sustained regulatory and litigation attention for exactly this reason.

Power cannot be stored, so it is traded as a sequence of markets rather than one

Electricity is the only major traded commodity that must be produced and consumed at the same instant. That single fact rewrites everything. There is no inventory to buffer a shock, no cost of carry to anchor the forward curve, and no way to arbitrage time by storing.

The market replaces storage with a sequence of markets that progressively refine the same delivery. Forward and futures markets trade months or years ahead. A day-ahead auction sets a price for each hour of the following day. Intraday markets let participants adjust as weather and demand forecasts change. A balancing or real-time market, run by the system operator, resolves the final gap between what was scheduled and what actually flowed.

Prices in power do things prices elsewhere cannot. They go negative when inflexible generation and subsidised renewable output exceed demand and it is cheaper to pay someone to consume than to shut a plant down. They spike by orders of magnitude for a single scarce hour. And they are location-specific to a degree oil never is, because transmission constraints mean the same megawatt-hour is worth different amounts at different points on the network.

Gas and LNG add seasonality, and a contractual option worth trading on its own

Natural gas is storable, which restores the cost-of-carry logic, but heating demand is strongly seasonal, so the forward curve repeats an annual shape rather than sloping smoothly. Storage is therefore not merely a cost. It is the instrument connecting summer injection to winter withdrawal, and its economic value is the summer-to-winter spread.

Liquefied natural gas adds the most interesting contractual feature in commodities: the destination clause. A cargo of LNG can physically go to any basin with a receiving terminal, so the contract must state whether it may be diverted. A cargo with destination flexibility is worth more than an identical cargo without it, because its holder can redirect it to whichever market is paying most. That flexibility is a real option, it is priced explicitly, and desks manage portfolios of it.

Agricultural markets are governed by biology and by politics

Grains and softs are produced in annual harvests, which makes supply inelastic within a season and splits the forward curve into old crop and new crop. Quality is measured in dimensions with no energy equivalent, such as protein and moisture in wheat, oil content in soybeans, or bean size and defect counts in coffee, and those measurements drive contractual premiums and discounts exactly as sulphur does in crude.

Two features dominate the risk. Weather is the primary supply variable, and it is forecastable over days while the exposure runs over months. And government intervention is routine rather than exceptional: export bans, tariffs, strategic reserves and subsidy regimes can move a market faster than any fundamental. A small number of large agricultural merchants dominate the physical trade, and their advantage comes from owning elevators, crushing plants and port terminals that generate information about real flows.

Family Structural quirk Delivery mechanism The risk that dominates
Crude and products Priced as a differential to an assessed benchmark, fixed after the trade Cargo, bill of lading, vessel Basis across grade, location and time
Base metals Daily prompt dates to three months rather than monthly contracts Warrant against approved warehouse stock Warehouse queues and the cost of financing stock
Power Unstorable; forward, day-ahead, intraday and balancing markets in sequence Scheduled delivery to a grid node Extreme short-horizon spikes and negative prices
Gas and LNG Strong seasonality; storage links summer to winter Pipeline nomination, or LNG cargo with destination terms Seasonal spread and the value of diversion optionality
Agriculture Annual harvest; old crop versus new crop Bulk vessel, elevator, warehouse receipt Weather and government intervention

The commodities desk sits inside markets, but it behaves like a supply chain business

A bank is many businesses, and this one is the odd cousin

An investment bank is a collection of businesses under one roof. Advisory bankers build long relationships with corporate clients and sell them transactions. The markets business, organised as Fixed Income, Currencies and Commodities alongside Equities, deals with investors and prices risk continuously. Commodities sits inside that markets division, usually as the C in FICC.

It is the odd cousin because it is the only markets business that may need a ship. A rates desk trading interest rate swaps needs a pricing model, a risk system and a clearing connection. A commodities desk running physical flow needs all of that plus a chartering capability, a scheduling team, an inspection contract and a letter of credit facility. The front office looks like trading. The back half looks like logistics.

The desk earns money in four distinguishable ways

Market making is the base layer. The desk quotes a two-way price to clients and earns the spread between where it buys and where it sells, managing the residual position it accumulates. This is flow business, it is repeatable, and it depends on having enough client volume to net one side against the other.

Hedging solutions are the client-driven layer. An airline wants to fix its jet fuel cost for the next eighteen months. A copper miner wants to lock in a floor price without giving up all the upside. The desk structures a swap, a collar or a series of options to deliver exactly that, and prices in the cost of hedging its own resulting exposure. Margins here are wider than in flow because the product is tailored.

Physical trading is the leg most banks have retreated from but not abandoned. Buying, shipping and selling actual cargoes captures the differential between locations, grades and time periods. It is capital-intensive, operationally demanding and reputationally exposed, and after the post-crisis reforms most banks narrowed it sharply. What survives is typically physical activity tied to client franchise, financing or a small number of markets where the bank has genuine infrastructure.

Financing and structured solutions are where the bank's balance sheet earns its keep. Lending against inventory, prepayment structures for producers, borrowing bases for traders and repo against warehouse warrants are all commodity businesses that look like credit. They are often the most reliably profitable part of the franchise because they are less dependent on volatility.

Independent trading houses beat banks at physical for structural reasons

It is worth being precise about why the physical business migrated rather than treating it as an accident of regulation. Trading houses own physical assets — terminals, blending facilities, vessels, storage — and those assets generate information about real flows that no screen provides. They operate under lighter prudential regulation, so capital deployed against inventory costs them less than it costs a bank. Their ownership structures tolerate concentrated positions and volatile earnings in a way a listed bank's shareholders do not. And their decision-making is built for speed rather than for committee.

The bank's advantages run the other way: a balance sheet, a global client franchise, credit expertise and derivatives capability. That is why the equilibrium settled where it did. Banks finance and hedge the physical market, trading houses move the molecules, and each is a major customer of the other.

Six products cover the market, and they differ in speed, complexity and who bears what

Under the commodities umbrella sit a small number of distinct product types. They share the same underlying reference prices but differ in how fast they trade, how much operational machinery they need and where the risk lands.

The exchange-traded future is the simplest and the fastest

A futures contract is a standardised agreement to buy or sell a fixed quantity of a specified grade at a fixed date and place, traded on an exchange and guaranteed by a clearing house. Standardisation is total: quantity, quality, delivery point and dates are all set by the exchange, and the only variable is price.

The clearing house is what makes it work. It interposes itself between buyer and seller so that neither has credit exposure to the other, and it collects margin daily to keep that guarantee good. Execution takes milliseconds and the operational load is close to zero. Most futures positions are closed before delivery; the contract is a price instrument, not a procurement method.

The OTC swap is the same economics without the standardisation

A commodity swap is a bilateral contract to exchange a fixed price for a floating one over a series of periods, settled in cash against a published benchmark. Nothing is delivered. It exists because real hedging needs rarely match exchange contract specifications: an airline burns fuel continuously, not in 1,000-barrel monthly blocks at a single delivery point.

Swaps are documented under an ISDA Master Agreement, the standard contract framework that governs bilateral derivatives, and since the post-crisis reforms many standardised commodity swaps are cleared through a central counterparty anyway. The tailoring is in the reference price, the volume profile and the tenor.

Options add asymmetry, and asymmetry is what clients usually want

An option gives the buyer the right but not the obligation to transact at a set price. A consumer buys a call to cap its cost while keeping the benefit of a fall. A producer buys a put to floor its revenue. A collar combines the two and can be structured to cost nothing upfront, which is why it is the most commonly executed corporate hedge in the market.

Options are where the desk's quantitative capability shows. Pricing them requires a volatility surface, and commodity volatility is badly behaved: it spikes on inventory reports, it is seasonal in gas and power, and it is not symmetric, because a supply disruption moves prices up far faster than a demand fall moves them down.

The physical spot cargo is where the process becomes a logistics problem

A spot cargo is a one-off purchase or sale of a specific physical parcel for near-term delivery. Everything the futures contract standardised away comes back: the exact grade, the loading window, the vessel, the delivery basis, the inspection regime, the payment terms. The trade is agreed in minutes and takes two to three months to complete.

This is the reference product for this module, and the next section explains why.

The physical term contract turns one cargo into a relationship

A term contract commits the parties to a recurring volume — a cargo a month for a year, or a fixed annual tonnage — at a formula price. Producers use term contracts to secure offtake and buyers use them to secure supply. The pricing formula is negotiated once and applied repeatedly, which makes the negotiation of that formula one of the highest-value activities in the market.

Structured and financing transactions combine the other five

Prepayment financings, inventory monetisations, tolling agreements and volumetric production payments bundle a physical flow with a credit facility and a hedge. They are the most complex products in the market by a distance, they involve legal, credit and trading teams simultaneously, and they take months to close.

Product Typical time to execute Operational complexity Where the desk's margin comes from Key technology
Exchange future Milliseconds to seconds Very low — cleared, no delivery Bid-offer spread, very thin Execution gateway, FIX, clearing interface
OTC swap Minutes Low — cash settled, confirmation only Spread over the hedgeable curve Pricing engine, confirmation platform, CCP link
Option / collar Minutes to hours Low to moderate Volatility spread and structuring premium Volatility surface, risk engine, Greeks
Physical spot cargo Minutes to agree, 2-3 months to complete Very high — vessel, inspection, documents Location, grade and timing differentials Full CTRM, scheduling, document management
Physical term contract Weeks to negotiate, years to run Very high and recurring Formula margin plus optionality in the terms CTRM with contract management, nomination workflow
Structured financing Months Extreme — legal, credit, trading, ops Credit spread plus embedded hedge margin CTRM plus credit systems plus deal-specific build

The most important trades are spreads between commodities, not bets on one

A processing spread is the margin of a transformation, and it trades as a single instrument

Many commodities are the input to another commodity. Crude oil becomes gasoline, diesel and jet fuel in a refinery. Natural gas becomes electricity in a power station. Soybeans become meal and oil in a crushing plant. Wherever that is true, the difference between the output value and the input cost is itself a tradeable quantity, and it is usually more useful to trade than either leg alone.

Figure 1 A processing spread is the margin of a transformation, traded as one position rather than two

The crack spread is the refining margin: the value of the products a barrel yields, minus the cost of the barrel. A refinery's economics depend on that difference and not on the absolute price of oil, so a refiner hedges by selling products and buying crude together, or more efficiently by trading the spread as one instrument.

The spark spread does the same for power generation: the price of electricity minus the cost of the gas needed to produce it, adjusted for the plant's efficiency, which is expressed as its heat rate. The dark spread is the coal equivalent. Where carbon allowances must be surrendered against emissions, the spread net of that cost is called the clean spark or clean dark spread, and the choice between running gas or coal plant is decided by whichever clean spread is higher. The crush spread is the agricultural version: the combined value of soybean meal and oil minus the cost of the beans.

Trading a spread changes the risk, and therefore changes the system

A spread is two or more legs held deliberately against each other, and it behaves like basis risk rather than like flat price. It is smaller, less volatile and considerably less liquid — precisely the profile that made Amaranth's calendar spread position fatal, as Part 5 describes.

For the systems this matters concretely. A spread must be represented as one economic position rather than as unrelated legs, or the risk report shows two large opposing exposures and the desk's actual risk becomes invisible. This is the same linkage problem as pairing a cargo with its hedge, wearing a different costume.

Storage and transport are options, and pricing them is the physical trader's real skill

An asset that lets you choose later is worth more than one that does not

A storage tank is not merely a cost centre. It gives its holder the right, but not the obligation, to sell today or in any future month. That is an option, and its value rises with volatility exactly as a financial option's does. This is why storage capacity leases at prices tracking expected volatility rather than the cost of the steel.

The same logic governs transport. A vessel or a pipeline slot is the right to move a commodity from one price to another. When the spread between two locations exceeds the cost of moving between them, the option is in the money and the trader exercises it by shipping. When it is not, the trader holds the option and waits.

Flexibility written into a contract is optionality too, and two examples have already appeared. The volume tolerance lets whoever holds it lift more when the cargo is worth more than the contract price and less when it is not. The LNG destination clause lets its holder redirect a cargo to the best-paying basin. Neither shows up as a line item on an invoice, and both are worth real money.

This is why physical traders think in options even when they never trade one

The habit worth taking from this section is to ask, of every physical asset and every contract clause, what choice it confers and when that choice must be made. A trader who values a storage lease at its rental cost will systematically underpay for volatility and overpay for certainty. A system that models a tolerance clause as a fixed quantity will misstate the position and the risk at the same time.

The physical crude cargo is the reference case, because it contains every other product inside it

One cargo exercises the entire machine

This module follows a physical crude oil cargo from origination to final settlement, because it is the only product that touches every stage, every team and every system. It has a price formula, so it teaches benchmarks. It has a flat-price exposure that must be hedged, so it drags the futures and swap products in with it. It has a vessel, so it teaches chartering, laytime and demurrage. It has a quality specification, so it teaches inspection and claims. It has a pricing period that falls after the deal, so it teaches actualisation. It has documents that control title, so it teaches letters of credit and the mechanics of getting paid.

A reader who understands a crude cargo end to end can pick up a copper cathode trade, an LNG cargo or a grain shipment in an afternoon. The vocabulary changes and the machine does not.

The unit of the trade is the parcel, and its size is set by the ship

Crude moves in parcels sized to vessel classes. A Very Large Crude Carrier holds roughly two million barrels. A Suezmax carries around one million and an Aframax somewhere between five hundred thousand and eight hundred thousand, depending on the grade's density. These are not arbitrary numbers; they are set by the physical constraints of canals, straits and terminal draughts, and they determine the minimum economic size of a trade.

This is a good early example of physical reality writing the rules of the financial product. You cannot trade a third of a cargo because a third of a cargo cannot be shipped on its own.

The market is enormous, and its size is measured in two incompatible ways

Physical volume and traded volume are different numbers, and the second is much larger

Global crude oil consumption runs at roughly one hundred million barrels a day. That is the physical market. The traded market is a large multiple of it, because the same barrel is bought and sold many times on paper before, during and after it physically moves, and because financial participants trade the price without ever touching the product.

The ratio of paper volume to physical volume — sometimes called the churn — is a measure of market liquidity, not of speculation gone wrong. A producer hedging output, a refiner hedging input, a trading house hedging a cargo and a fund taking the other side of all three each generate paper volume against a single physical barrel. Without them, the producer would have no one to sell risk to.

A handful of benchmarks anchor every price in the world

Almost no physical commodity trades at a freely negotiated absolute price. It trades at a differential to a published benchmark. In crude, the dominant references are Brent for the Atlantic basin and much of the world's term pricing, West Texas Intermediate for North America, and Dubai for barrels flowing east. In natural gas, Henry Hub anchors North America and TTF anchors continental Europe, with JKM serving as the Asian LNG marker. Base metals price off London Metal Exchange settlements, and grains off Chicago futures.

Benchmark Commodity Region it anchors Price source type
Dated Brent Crude oil Atlantic basin and much of global term pricing PRA assessment of physical cargoes
WTI Crude oil North America Exchange futures, physically deliverable
Dubai Crude oil Barrels flowing east PRA assessment
Henry Hub Natural gas North America Exchange futures, physically deliverable
TTF Natural gas Continental Europe Exchange and OTC traded hub price
JKM LNG Asia PRA assessment
LME settlements Base metals Global Exchange, physically deliverable against warrants

Two kinds of institution produce these numbers, and confusing them is a common error. Exchanges such as ICE and CME produce prices as a by-product of trading: the number is whatever the last trade printed. Price Reporting Agencies such as Platts and Argus produce assessments by surveying and observing the physical market and applying a published methodology, because much of the physical market does not trade on any exchange at all.

The distinction matters commercially. An exchange price is an outcome. A PRA assessment is a judgement made against a methodology, most visibly in the structured daily window during which bids, offers and trades are collected and assessed. Because trillions of dollars of contracts settle against those assessments, behaviour inside that window is closely watched by regulators and tightly controlled inside banks.

Fees in commodities are margins, not percentages

Unlike an equity offering with a stated underwriting fee, a commodities desk rarely charges an explicit fee. It earns a margin embedded in the price. On a liquid futures trade the spread might be a fraction of a cent per unit. On a tailored corporate hedge it might be a few cents. On a physical cargo the desk's economics are the difference between the buying and selling differentials, less freight, financing, insurance and any demurrage it eats.

That has a consequence worth stating plainly: profitability in commodities is invisible on the ticket. You cannot read a trade and know what the desk made. Only the position, valued correctly against the right curve, tells you — which is why the risk and product control functions in Part 3 carry more weight here than in almost any other business.

Paper and wet are two markets that must be understood as one

The paper market prices risk; the wet market moves molecules

Traders speak of paper barrels and wet barrels. Paper is the financial layer: futures, swaps and options, cash-settled, cleared or bilateral, executed in seconds. Wet is the physical layer: real cargoes with real vessels and real documents.

The two are joined by the pricing formula. A physical cargo priced against Dated Brent is, economically, a paper Brent position plus a differential. That is why the desk can buy a cargo and immediately sell futures against it: the futures neutralise the benchmark exposure and leave the desk holding only the differential, the freight and the timing. This decomposition is the single most important mental move in commodity trading, and Part 2 walks through it with numbers.

Cash settlement and physical settlement are choices, not properties

A common misunderstanding is that futures are financial and cargoes are physical. The cleaner statement is that every contract specifies a settlement mechanism, and delivery is one option among several. WTI futures are physically deliverable at Cushing, Oklahoma. Brent futures cash-settle against an index. LME contracts are physically deliverable against warehouse warrants and a meaningful share do go to delivery. The choice of settlement mechanism drives an enormous amount of downstream system design, because a deliverable contract needs a delivery process even if it is rarely used.

The energy transition is rewriting what the desk trades

Carbon has become a commodity with a market of its own

An emissions allowance is a permit to emit one tonne of carbon dioxide. Under a cap-and-trade system such as the European Union Emissions Trading System, a cap is set, allowances are issued and emitters must surrender enough to cover what they emit. Because allowances trade, a price emerges, and that price enters the economics of every business that burns fuel.

For a commodities desk this is not a side market. The carbon price is a direct input to the clean spark and clean dark spreads described above, which means it helps decide whether gas or coal plant is dispatched, and therefore drives the demand for both fuels. Carbon is now a variable inside the fundamentals of commodities that have been traded for a century.

The transition adds new products and destabilises old ones

Power purchase agreements with renewable generators, guarantees of origin certifying that electricity came from a renewable source, biofuels and the mandates that create demand for them, and developing markets in hydrogen and voluntary carbon credits are all now traded or structured by commodity desks. Each carries its own origin and certification problem, and those land directly on the reference data and document management systems described in Part 4. Proving that a molecule or an electron came from where the certificate claims is a data lineage problem before it is a commercial one.

The transition also makes the older markets more volatile rather than less. A generation fleet with a large intermittent share needs more flexibility, and flexibility — storage, fast-start generation, optionality — is exactly what a commodities desk sells.

The business is violently cyclical, and the cycle drives the technology budget

Volatility is the desk's raw material, and it does not arrive evenly

A commodities desk makes money from movement and from dislocation. In a calm, well-supplied market, differentials are tight, clients see no urgency to hedge and volumes fall. In a disrupted market, differentials blow out, every consumer wants protection and volumes surge. Revenue is therefore not a smooth function of economic activity; it is a function of uncertainty.

That makes the business genuinely cyclical in a way few others are. Headcount, risk limits and system investment expand after a volatile, profitable year and contract after a quiet one, often just before the next disruption arrives.

The shape of the forward curve is itself a trading signal

The forward curve plots the price of a commodity for delivery at successive future dates. Its shape has a name and an economic meaning.

Figure 2 Contango and backwardation: the two shapes of the forward curve and what each one pays you to do
Contango means later is more expensive

When forward prices sit above the spot price, the market is in contango. This typically signals ample supply: nobody needs the commodity urgently, and the forward premium reflects the cost of carrying it. Contango rewards storage. If the premium for delivery six months out exceeds the cost of tank rent, insurance and financing, a trader can buy physically, store, sell forward and lock the difference. In the extreme contango of 2020, this trade absorbed so much capacity that traders chartered tankers purely as floating storage.

Backwardation means sooner is more expensive

When spot sits above forward prices, the market is backwardated. This signals scarcity: buyers are paying a premium to have the commodity now rather than later. Backwardation punishes storage and rewards holding inventory. It also creates a positive roll yield for a long futures position, because each expiring contract is replaced by a cheaper one further out.

The recent cycle has been the most extreme in a generation

The period from 2020 onward compressed several full cycles into a few years. In April 2020, demand collapse combined with storage saturation drove the expiring WTI futures contract to settle below zero for the first time in history, at roughly minus thirty-eight dollars a barrel. Holders of long positions with no capacity to take delivery discovered that a physically deliverable contract is a physical obligation.

The recovery and the 2022 European energy crisis pushed prices and volatility to the other extreme, and gas and power markets in Europe repriced by an order of magnitude. Margin calls on hedged positions reached levels that threatened the solvency of otherwise sound utilities, because a hedge that is winning economically still consumes cash when it is losing on a daily mark. In March 2022, the London Metal Exchange suspended nickel trading and cancelled a session's trades after a short squeeze drove prices to levels that would have bankrupted clearing members.

Each of these episodes taught the same lesson from a different angle: the paper and physical layers are joined by cash, and cash is where crises actually arrive.

The cycle drives the IT roadmap directly. After a volatile year, investment flows to risk systems that can reprice faster, margin systems that can forecast collateral demands under stress, and connectivity that can absorb ten times the normal message volume. After a quiet year, projects consolidate platforms and cut vendor cost. The 2020-2022 period pushed three specific capabilities up every commodity bank's priority list: intraday risk rather than overnight batch, forward-looking margin and liquidity forecasting rather than backward-looking reporting, and stress testing that models a hedge winning on paper while draining cash.

  1. What does fungibility mean, and why is it the precondition for a traded commodity market?

  2. Name the three properties that separate commodities from financial assets, and give one consequence of each.

  3. What is the difference between an exchange price and a PRA assessment?

  4. A desk buys a physical cargo priced against Dated Brent and immediately sells Brent futures. What risk has it removed and what does it still hold?

  5. What does contango pay a trader to do, and what does backwardation pay a trader to do?

  6. Why is desk profitability invisible on an individual physical trade ticket?

  7. Why can electricity prices go negative when oil prices essentially cannot?

  8. What is a crack spread, and why does a refiner prefer to trade it rather than crude alone?

  9. In what sense is a storage tank an option, and what follows for how it should be priced?

  10. Give two structural reasons independent trading houses out-compete banks in physical trading.

  11. What did the April 2020 negative WTI settlement demonstrate about physically deliverable contracts?

1

  1. Why is fungibility the precondition for a traded commodity market?

  2. Which property makes freight part of the price?

  3. What is a PRA assessment?

  4. A desk buys a cargo priced on Dated Brent and sells Brent futures. What does it still hold?

2The Trade Lifecycle End to End

This is the spine of the module. One trade, followed from the market view that prompts it to the day the final invoice clears, with every system and every decision named along the way. Read this part slowly; the rest of the module takes it apart.

Nine stages carry a commodity trade from idea to cash

The lifecycle is long, and most of it happens after the price is agreed

In equities or rates, the trade is the event and everything after it is administration. In commodities the trade is the beginning. Agreeing to buy a cargo takes two minutes on the phone. Completing that purchase takes two to three months, involves a vessel, an inspector, a bank guarantee and several thousand data fields, and the price is not even known until most of it is over.

Stage Typical duration Core output Who leads
1. Pre-trade and price discovery Continuous A price the desk is willing to show Trading, with research and credit
2. Execution and trade capture Minutes A booked deal ticket in the CTRM Trading and middle office
3. Confirmation and documentation T+0 to T+1 A matched confirmation, contract terms agreed Middle office and legal
4. Hedging and risk transfer Same day A linked financial position neutralising flat price Trading and risk
5. Logistics and scheduling Days to weeks A nominated vessel and an agreed loading window Operations and chartering
6. Delivery and inspection The loading and discharge windows Bill of lading and inspection certificates Operations and inspectors
7. Pricing actualisation The pricing period, often after loading A final, fixed price per unit Middle office and product control
8. Invoicing and settlement 30 to 60 days from bill of lading Cash received against documents Operations, finance and trade finance
9. Post-trade and lifecycle To expiry and beyond Reconciled position, P&L, regulatory reports Product control, risk and reporting

Effort and risk peak twice, not once

Most deal processes have a single moment of maximum pressure. Commodities has two, and understanding why is the fastest way to understand the business.

The first peak is execution. Prices move continuously, a quote is live for seconds, and the decision to commit is made in the time it takes to say yes. Everything about the trade — grade, volume, formula, delivery basis, counterparty — is fixed in that moment, and any error in capturing it propagates through every subsequent stage.

The second peak arrives weeks later, around delivery and price fixing. The vessel must arrive within its window, the quantity and quality must be certified, and the price must be calculated from published assessments over a defined period. This is where a trade that looked clean on the ticket can lose money to a three-day berthing delay, a sulphur reading outside tolerance or a benchmark that moved during the pricing window.

Between those peaks sits a long plateau of operational work that is unglamorous, entirely necessary, and where most of the headcount actually is.

Figure 3 The commodity trade lifecycle: nine stages, two peaks of risk and effort, and a long operational plateau between them

Step 1 — Price discovery happens continuously, and the desk is quoting long before anyone trades

A commodity price is assembled, not looked up

Before the desk can show a price on a cargo it must know what that cargo is worth, and no screen displays that number. The desk builds it. It starts from the benchmark, which is observable. It adds the differential for the specific grade, which comes from recent comparable deals, broker indications and the PRA assessments. It adjusts for quality against the benchmark's own specification. It adds or subtracts freight to move the cargo between the two delivery bases. Finally it accounts for time, because a cargo loading in six weeks is not the same asset as one loading tomorrow.

Figure 4 Building a delivered cargo price: benchmark plus differential, adjusted for quality, plus freight
The benchmark is the anchor and the only piece everyone agrees on

Dated Brent, in this example, is the reference. It is a physical assessment of North Sea cargoes loading within a defined forward window, published daily by a price reporting agency. Because it is physical rather than a futures settlement, it reflects the actual market for prompt cargoes, which is exactly why term contracts across the world reference it.

The differential is where the desk's view lives

The differential is what a specific grade is worth relative to the benchmark, and it is the number the desk is genuinely trading. It moves with refinery demand for that particular crude's yield, with freight economics, with competing grades and with supply disruptions. A trader who is bullish on diesel cracks is, in practice, bullish on the differentials of crudes that yield a lot of diesel.

Freight converts one delivery basis into another

Delivery bases are defined by Incoterms, the standard trade terms published by the International Chamber of Commerce. Two matter most here. Under FOB, free on board, the seller's obligation ends when the cargo passes into the vessel at the load port, and the buyer arranges and pays for shipping and bears the risk in transit. Under CFR, cost and freight, the seller arranges and pays the freight to the destination. CIF adds insurance to the seller's obligations.

Incoterm Who arranges and pays freight Who insures Where risk passes to the buyer
FOB — free on board Buyer Buyer On loading at the load port
CFR — cost and freight Seller Buyer On loading at the load port, despite the seller paying freight
CIF — cost, insurance and freight Seller Seller On loading at the load port
DAP — delivered at place Seller Seller On arrival at the named destination

The difference between an FOB price and a CFR price is, in principle, just the freight. In practice it is also a position: whoever charters the vessel owns the freight risk, and freight is itself a volatile traded market with its own rates, quoted for tankers on the Worldscale system.

A second method values the cargo from what it becomes, and the two must agree

A price built one way is an assertion. A price confirmed two ways is a valuation. Professional desks never rely on the build-up alone, because it inherits whatever the differential market happens to be quoting and gives no independent check on whether that quote makes sense.

The netback works backwards from the products the barrel becomes

The build-up starts at the benchmark and adds. The netback starts at the other end and subtracts. It asks what the cargo is worth to the refinery that will process it: take the value of the products a barrel of this grade yields, subtract the cost of refining it, and the remainder is the most a rational refiner could pay and still make money.

The yield pattern comes from the assay. A light sweet grade yields more of the valuable middle distillates and less heavy fuel oil, which is precisely why it commands a premium over a heavy sour grade.

Triangulation is the discipline of making three methods agree

Three independent estimates now exist. The build-up says the cargo costs the desk $84.90 delivered. Recent comparable transactions and the assessed differential put similar Mediterranean cargoes at Brent plus $4.40 to $4.80, or roughly $86.40 to $86.80. The netback puts the buyer's ceiling at $88.40.

Figure 5 Three independent methods converging on a defensible range

The sale at Brent plus $4.60, or $86.60, sits above the desk's cost, inside the comparable range and below the buyer's ceiling. That is a defensible price, and defensible is the operative word: it is the price the desk can justify to its own product control function, to the client, and to a regulator asking why this trade was done at this level.

When the three methods disagree sharply, that disagreement is information rather than an error. A netback far above the market differential says refining margins are strong and the differential should widen. A netback far below says the grade is overpriced and someone is about to stop buying it. Traders watch the gap between the methods more closely than they watch any single one of them.

Clients reach the desk in three distinguishable ways

Request for quote is the dominant mechanism for tailored business. A client asks the desk to price a specific structure — a year of monthly jet fuel swaps, a cargo of a named grade — and the desk responds with a firm price, live for a stated period. The desk is competing against a small number of other banks and trading houses, and it knows it.

Streaming prices serve the liquid, standardised end. For benchmark futures and vanilla swaps, the desk publishes continuously updated two-way prices through electronic channels, and clients execute against them without conversation. Volume is high, margin is thin, and the entire economics depend on automation.

Axes and market colour drive the physical business. The desk broadcasts what it is looking to buy or sell — its axes — to clients and brokers, because a physical cargo needs a specific counterparty who wants that grade, in that window, at that port. Physical trading is a matching problem before it is a pricing problem.

The assessment window is a controlled environment, not a free-for-all

Because so many contracts settle against PRA assessments, the structured daily window in which those agencies observe bids, offers and trades is one of the most closely supervised periods in the market. Trading during it is legitimate and necessary — the assessment needs real activity to observe. Trading during it in order to move the published number is market manipulation, and it has produced enforcement actions and criminal convictions.

Banks control this with hard rules: named authorised traders, pre-approval of window activity, full communication capture, and post-hoc surveillance comparing window behaviour to the desk's underlying position. Part 4 covers the systems that enforce it.

Pre-trade pricing is a data problem before it is a judgement problem. The desk needs live benchmark futures, daily PRA assessments, freight rates, forward curves for every grade it trades, quality assays, port and terminal restrictions, and the current inventory position — all in one place, all timestamped, all reconciled. This is the pricing and analytics layer, and it is typically part vendor market-data feed and part proprietary curve engine. The curve engine is almost always built in-house, because how a bank interpolates a differential curve between two observable points is genuinely proprietary and is the difference between a good price and a bad one.

Step 2 — Execution commits the bank, and trade capture is where the commitment becomes data

The trade is agreed in minutes and lives for months

Physical cargoes are still frequently agreed by voice or chat, between people who have dealt with each other for years, on terms compressed into a few lines. The recap message that follows — grade, quantity and tolerance, loading window, delivery basis, pricing formula, pricing period, payment terms, applicable general terms and conditions — is the commercial agreement in embryo. Everything downstream is an elaboration of it.

The speed is not carelessness. Differentials move, cargoes are scarce, and a counterparty who takes an hour to respond loses the cargo. But it means the moment of maximum commitment is also the moment of minimum documentation, and closing that gap is what the next two stages exist to do.

Trade capture is the single most consequential data event in the lifecycle

Trade capture is the act of recording an agreed trade into the bank's system of record, and everything downstream reads from it. The system is the Commodity Trading and Risk Management platform, universally called a CTRM, or ETRM when energy-specific. It is the spine of the entire architecture, and Part 4 dissects it.

A physical commodity deal ticket is not a simple object. Where an equity trade needs perhaps a dozen fields, a physical cargo needs well over a hundred, and each one drives a downstream consequence.

Field group Examples What it drives downstream
Instrument and grade Commodity, grade, quality specification, assay reference Valuation curve, quality claims, blending options
Quantity Volume, unit of measure, tolerance (e.g. +/- 5%), operational tolerance holder Invoice value, hedge ratio, vessel sizing
Pricing Benchmark, differential, pricing period, pricing basis, rounding, conversion factors Actualisation, P&L, hedge instrument selection
Delivery Incoterm, load port, discharge port, laycan window, title transfer point Freight cost, risk transfer, insurance, scheduling
Counterparty Legal entity, LEI, credit line, netting agreement, guarantor Credit check, sanctions screening, netting, reporting
Commercial terms Payment terms, credit instrument, demurrage rate, laytime allowance Settlement, financing cost, demurrage exposure
Internal Book, strategy, trader, hedge linkage, cost centre P&L attribution, risk aggregation, limit consumption
The pricing formula is the field that most often destroys value when it is wrong

A cargo priced on "Dated Brent plus $1.20, five days around bill of lading" and a cargo priced on "Dated Brent plus $1.20, the calendar month of delivery" are different trades with different risks and different hedges, and they look almost identical on a ticket. Mis-capturing the pricing period is one of the most expensive routine errors in the business, because the hedge will be placed against the wrong dates and the mismatch only becomes visible when the price is fixed weeks later.

The physical and financial legs must be explicitly linked

The cargo purchase and the futures sale that hedges it are two separate transactions in two separate markets, often booked in different modules of the CTRM. If the system does not carry an explicit link between them, the risk report shows a large outright long in physical and a large outright short in paper rather than a hedged position. Risk managers then see phantom exposure, limits are consumed incorrectly, and — far worse — if someone closes one leg without the other, a genuinely hedged position becomes an outright bet with nobody having decided to take it.

Figure 6 Two legs, one economic position: unless the CTRM links them, the hedge is invisible to risk

Step 3 — Confirmation turns a conversation into an enforceable contract

The recap is not the contract, and the gap between them is real risk

Confirmation is the process of both parties independently agreeing, in writing, the full terms of what they just traded. For physical cargoes it produces a contract incorporating a set of general terms and conditions — often the seller's standard GTCs, sometimes an industry standard set — covering title, risk transfer, measurement, claims, force majeure, sanctions and governing law. For financial derivatives it produces a confirmation under an ISDA Master Agreement.

Until confirmation is matched, the bank has an economic exposure recorded against terms that have not been mutually agreed. Disputes discovered at this stage are cheap. Disputes discovered when a vessel is at the berth are not.

The documentation stack has three layers, and each does a different job

The master agreement sets the rules for the relationship. An ISDA Master governs all derivatives between the two parties: netting, events of default, collateral through the Credit Support Annex. It is negotiated once and covers thousands of trades.

The general terms and conditions set the rules for physical trading. They cover measurement standards, tolerance, demurrage, inspection, title and risk transfer, and increasingly sanctions and vessel-eligibility clauses that have grown substantially in recent years.

The trade confirmation sets the terms of the individual deal. It is short, because it inherits everything else from the two layers above.

Confirmation is a matching problem, and the industry solves it differently on each side. Financial derivatives are confirmed electronically in FpML, an XML standard for derivative terms, through platforms that match both sides' submissions automatically and flag breaks. Physical contracts are still substantially document-based, exchanged as PDFs and negotiated by email, which is why confirmation backlogs are a physical-commodities problem far more than a derivatives one. An unconfirmed physical trade sitting for two weeks is a legal exposure and, in most banks, a reportable operational risk item.

Step 4 — Hedging separates the risk the desk wants from the risk it does not

The decomposition is the central intellectual move of the business

A physical cargo contains two economically distinct risks, and the desk wants only one of them. It wants the differential: its view that Kestrel Blend is cheap relative to Brent, and that moving it to the Mediterranean captures a location spread. It does not want the flat price: an outright bet on whether crude oil goes up or down, which is far larger and has nothing to do with the desk's expertise.

The cargo is worth roughly $170 million. A five-dollar move in the crude price changes its value by $10 million. The entire differential margin is $2.6 million. Unhedged, the trade's outcome is decided almost entirely by a variable the desk has no view on.

So the desk sells futures. Two million barrels of exposure, hedged with roughly two thousand ICE Brent futures contracts of a thousand barrels each. If crude falls ten dollars, the cargo loses $20 million and the futures gain approximately the same. What survives is the differential — which is what the desk was actually trying to trade.

What survives the hedge is basis risk, and basis risk is the real job

Basis risk is the risk that the hedge and the exposure do not move together. It never disappears; hedging converts a large, simple risk into a smaller, more complex one. In this trade, several basis risks remain.

The grade is not Brent. Kestrel Blend's differential can widen or narrow independently of the benchmark, and that is precisely the position the desk chose to hold.

The dates do not match perfectly. The physical prices over five days around the bill of lading; the futures contract expires on its own schedule. The desk must roll or adjust its hedge so the two align, and any gap between them is exposure.

The location is different. The cargo is worth what it is worth in the Mediterranean; the benchmark reflects the North Sea. Freight rates moving between trade date and shipment change the economics with no hedge in most cases, because freight derivatives are far less liquid than crude derivatives.

The volume is uncertain. The contract has an operational tolerance — say plus or minus five per cent — and the actual bill of lading quantity is only known at loading. A hedge sized to exactly two million barrels is wrong by up to a hundred thousand barrels the moment the vessel finishes loading.

Two risk measures you will meet on every desk

Value at Risk, universally shortened to VaR, states the loss a portfolio should not exceed over a given period at a given confidence level. A one-day 99 per cent VaR of ten million dollars means that on 99 days out of 100 the portfolio should lose less than that. It compresses a complicated position into one number, which is its usefulness and its danger: it says nothing about how bad the remaining one day in a hundred can be, and commodity markets specialise in exactly that tail.

The Greeks are the sensitivities of a position to the things that move it, named after the Greek letters conventionally used for them. Delta is the change in value for a one-unit move in the underlying price, and it is what tells the desk how many futures to sell against a cargo. Gamma is how fast delta itself changes. Vega is sensitivity to volatility, and theta is sensitivity to the passage of time. A desk running options lives by these numbers; a desk running only physical and futures mostly needs delta.

The hedge, demonstrated rather than asserted

Numbers make this concrete. Suppose Brent falls from $82.00 to $72.00 between the trade date and the pricing window, and the desk is holding the cargo unsold.

Position Value at $82 Brent Value at $72 Brent P&L
Physical cargo, 2m bbl, worth Brent + $1.20 $166.4m $146.4m -$20.0m
Short 2,000 Brent futures sold at $82.00 — — +$20.0m
Net outcome ~$0

The hedge did its job. A twenty-million-dollar swing in the value of the cargo was neutralised, and the desk's result no longer depends on the direction of crude oil.

Now add the complication that actually matters. Suppose the grade differential also moved, narrowing from plus $1.20 to plus $0.95 because Mediterranean refining demand softened.

Position Value at $82, diff +$1.20 Value at $72, diff +$0.95 P&L
Physical cargo, 2m bbl $166.4m $145.9m -$20.5m
Short 2,000 Brent futures — — +$20.0m
Net outcome -$0.5m

The residual is exactly twenty-five cents a barrel across two million barrels. That half-million-dollar loss is not a hedging failure; it is the position the desk deliberately took. It backed its view on the differential, and the differential went against it.

Clearing and margin turn a hedge into a cash obligation

The futures hedge is cleared through a central counterparty, which removes credit risk and replaces it with a daily cash obligation. Initial margin is posted upfront as a buffer. Variation margin then flows every day to reflect the position's mark-to-market move.

Here is the asymmetry that has broken real institutions. If crude prices rise, the futures short loses money and the desk pays variation margin in cash, daily. The physical cargo has gained an offsetting amount — but that gain is unrealised and will not turn into cash until the cargo is delivered and paid for, perhaps two months later. The position is economically hedged and cash-negative at the same time.

Figure 7 Economically hedged and cash negative at the same time: the funding gap that has broken real institutions

That mismatch is a funding risk, it scales with volatility, and it is the mechanism behind the 2022 European energy margin crisis and the Metallgesellschaft failure discussed in Part 5. A desk must fund its hedges as well as place them.

The internal gauntlet: what the bank checks before it lets any of this happen

Three gates stand between a trader's idea and a live commitment

A trader cannot simply agree a $170 million cargo purchase because the price looks good. Three internal gates must already be open, and the technology that enforces them is as important as the technology that prices the trade.

Figure 8 Four gates stand between a trader's idea and a live commitment
The counterparty gate: onboarding, KYC and sanctions

Before any trade, the counterparty must be onboarded: a legal entity identified, beneficial ownership established, know-your-customer checks completed, and sanctions screening passed. In commodities this screening is unusually broad, because sanctions attach not only to counterparties but to cargo origin, vessel, vessel owner, flag state, ship-to-ship transfer history and even the route taken.

A cargo can be legally purchased from a sanctioned-adjacent origin, loaded onto a vessel that later appears on a designation list, and become an enforcement problem after the trade. Screening therefore runs continuously through the life of the trade, not once at onboarding. This is a genuine technology requirement: vessel and cargo screening integrated into the scheduling system, re-run on every change of nomination.

The credit gate: does the bank have room for this counterparty?

Credit must approve an exposure limit before the trade, and the exposure being measured is not the notional. It is the potential future exposure — how much the bank could lose if the counterparty defaults and the position must be replaced at market prices, modelled across possible price paths. That number moves with volatility, which means a limit that was comfortable in a calm market can be breached by a price spike with no new trading at all.

For physical trades, credit exposure has a distinctive shape. Between the bill of lading and payment, the bank has typically parted with the cargo and not yet been paid — a straightforward receivable exposure often mitigated by a letter of credit. Before delivery, the exposure is performance risk: the risk the counterparty simply fails to deliver or lift, forcing replacement at a worse price.

The mandate gate: is the desk allowed to do this at all?

New products, new markets, new jurisdictions and new physical activities pass through governance committees before the first trade. A new product approval process examines whether the bank can price it, book it, risk-manage it, confirm it, settle it, report it and unwind it. In practice, the question that kills more proposals than any other is whether the CTRM can represent the product without manual workarounds, because a product that lives in a spreadsheet is invisible to risk.

Physical commodity activity carries an additional reputational and environmental review that financial products do not. Cargo origin, environmental exposure, counterparty conduct history and the jurisdictions involved are assessed by a committee that can veto a commercially attractive trade on non-commercial grounds.

Gate The question it answers Who owns it What it blocks if it fails
Counterparty Do we know who this is, and is any part of the trade sanctioned? Onboarding and compliance Any dealing at all, and it can stop a cargo mid-voyage
Credit Do we have room for this exposure if the counterparty defaults? Credit risk The size of the trade, or the trade entirely
Mandate Are we approved to trade this product, and can we actually book, risk-manage and settle it? New product governance The product, until the operating model exists
Reputational and environmental Should we do this even though we may? Reputational risk committee A commercially attractive trade, on non-commercial grounds

Step 5 — Scheduling turns a contract into a movement, and this is where commodities stops resembling finance

Operations is not back office here; it is the business

Scheduling is the function that converts contractual obligations into an executable plan of vessels, dates and berths. The schedulers hold a rolling view of every cargo the desk has bought and sold, every vessel available, every loading window promised and every discharge slot booked, and they solve a constraint problem continuously as those inputs change.

They are not administrators processing the trader's decision. A trader who agrees a cargo with a loading window the schedulers cannot serve has created a loss, and in well-run desks the scheduler is consulted before the trade rather than informed after it.

Nomination is the formal mechanism that fixes the plan

Nomination is the contractual act of declaring the specific details that the contract left open. The buyer under an FOB contract nominates the vessel it will send, by a deadline stated in the contract. The seller nominates the loading terminal and confirms the three-day loading range. Each nomination is a formal notice with legal effect, and each has a deadline that carries consequences if missed.

Nomination is also the commodities equivalent of allocation. When a desk holds several cargoes of a grade and several contracts requiring that grade, it decides which physical parcel satisfies which contractual commitment. Those decisions are not neutral. Cargoes differ in quality, loading date and cost, and contracts differ in price and tolerance. Matching the cheapest cargo to the best-priced contract is real money, and the optionality embedded in that matching is a genuine source of value the trader priced into the deal.

The tolerance is an option, and someone owns it

Physical contracts specify a quantity with a tolerance — five per cent either way is typical — and they state who chooses within that range. This is not administrative flexibility. It is a valuable option: whoever holds it can lift more when the cargo is worth more than the contract price and less when it is worth less. Traders negotiate it deliberately, and a system that does not capture who holds the tolerance will misprice the contract and mis-hedge the volume.

Chartering brings a second traded market into the trade

Under CFR or CIF terms the seller must move the cargo, and moving it means chartering a vessel. Freight is its own market with its own brokers, its own rate structure and its own volatility. A voyage charter hires a specific vessel for a specific voyage; a time charter hires it for a period. Rates for tankers are quoted on the Worldscale system, a published schedule of nominal rates for standard voyages against which the market trades a percentage.

Freight moved dramatically during the 2022 disruption of trade routes, and a desk that fixed a cargo price before fixing its freight discovered that a differential margin of a dollar a barrel can be entirely consumed by a freight market that repriced between the two decisions.

Freight is itself hedgeable, though far less cleanly than crude

Because freight is a volatile traded market, it can be hedged. Forward Freight Agreements are cash-settled contracts that pay the difference between an agreed rate and a published freight index for a defined route and period. A desk that has sold a cargo on a delivered basis but not yet fixed a vessel can use them to lock its shipping cost.

The caveat is liquidity. Freight derivatives are thinner than crude derivatives by a wide margin, they cover standard routes rather than the specific voyage a cargo actually needs, and the index they settle against may diverge from the rate the desk ends up paying. Hedging freight therefore swaps a large exposure for a smaller basis, exactly as hedging crude does — and the residual is proportionally larger. Many desks accept freight risk unhedged on short-dated cargoes for precisely this reason, which is a decision rather than an oversight, and it should be a documented one.

Blending turns off-specification components into an on-specification cargo

One of the least visible and most reliable sources of physical trading profit is blending: combining two or more streams so that the mixture meets a specification neither component meets on its own, or so that a valuable component is not wasted delivering quality the buyer does not pay for.

If a contract permits sulphur up to a stated limit, delivering material well below that limit is giving away quality for free. Blending a cheaper, higher-sulphur stream up to just inside the tolerance captures that giveaway as margin. The same logic applies to density, vapour pressure, and in metals and agriculture to purity, protein and moisture.

This is a constrained optimisation problem — maximise the value of the blend subject to every quality parameter staying within contractual limits — and it is one of the few places in commodity trading where a genuinely mathematical technique produces money directly. Part 4 returns to it as a systems capability.

Vessel vetting is a compliance function as much as a safety one

Before a vessel is accepted it must be vetted: age, class, inspection history, insurance, ownership structure, flag, and sanctions status of every party in the chain. The compliance dimension has expanded significantly, and vetting now routinely includes tracking whether a vessel has disabled its transponder or conducted undisclosed ship-to-ship transfers.

Scheduling and chartering run on systems that most finance technologists have never seen: vessel tracking through AIS satellite feeds, port and terminal reference databases holding draught and berth restrictions, laytime calculators, and demurrage engines. These integrate with the CTRM but are frequently separate platforms, and the interface between them is one of the most break-prone in the whole architecture. When the schedule changes in the operations system but the CTRM still holds the original dates, the pricing period is wrong, the hedge is misaligned and the exposure report is stale — and nobody notices until actualisation.

Step 6 — Delivery is the irreversible moment, and the documents matter more than the cargo

Figure 9 The physical journey, and where title and risk actually pass under each Incoterm

Loading creates the two documents that govern everything after it

When the vessel berths, the master issues a Notice of Readiness, and the contractual clock starts. Cargo is pumped aboard over one to two days, and two documents are produced that dominate the rest of the lifecycle.

The bill of lading is the document of title. Issued by the carrier when the cargo is loaded, it evidences receipt, states the quantity and date, and — critically — whoever lawfully holds the original bill of lading controls the cargo. It is a negotiable instrument. It can be sold, endorsed and pledged, which is why it sits at the centre of both trading and trade finance. Its date also usually anchors the pricing period, which means the bill of lading date is simultaneously a legal, logistical and pricing event.

The inspection certificate establishes quantity and quality. An independent inspector, appointed jointly and paid by both parties, measures the volume loaded and takes samples that are analysed against the contractual specification. Their certificate is normally binding on both sides. This is the mechanism that makes a fungible commodity contract enforceable: without an agreed independent measurement, every cargo would end in a dispute.

The letter of indemnity exists because ships move faster than paperwork

Original bills of lading travel through banks and are regularly still in transit when the vessel arrives. Since the carrier may not discharge without them, the buyer issues a letter of indemnity — a promise to indemnify the carrier for discharging without presentation of the original documents. It is routine, it is essential, and it is a real contingent liability that must be tracked, because the bank has guaranteed something against documents it does not yet hold.

Laytime and demurrage convert delay directly into cash

The contract allows a fixed period — laytime — for loading or discharging, typically counted in running hours from the Notice of Readiness. If operations exceed it, the charterer pays demurrage to the shipowner at an agreed daily rate. If they finish early, some contracts pay despatch in the other direction.

Demurrage is not a rounding error. On a VLCC, daily rates run into the tens of thousands of dollars, and delays of several days are common when berths are congested or documents are late. On a trade with a $1.30 per barrel gross margin, a few days of demurrage takes a visible bite out of the outcome. Demurrage claims are calculated from the vessel's statement of facts, are frequently disputed, and can remain open for months after the cargo has been delivered and paid for.

Step 7 — The price is fixed after the trade, and this is what makes commodities strange

Figure 10 The desk is committed for weeks before the price exists

Actualisation is the moment the formula becomes a number

Actualisation is the process of calculating the final price by applying the contractual formula to published data once the pricing period has passed. Until then, the cargo has a provisional value; afterwards it has a price.

In the running case, the formula is Dated Brent plus $1.20, averaged over the five days around the bill of lading date. So the desk collects the published Dated Brent assessments for those five specific days, averages them, adds the differential, and applies any quality adjustment. Only then does the invoice value exist.

The consequences of this are worth stating slowly. The desk agreed to buy the cargo perhaps six weeks before the price was known. It knew the differential and the formula; it did not know the flat price. That is exactly why the futures hedge is placed against the pricing period rather than the trade date — the hedge must be live during the days that determine the price, not during the days the deal was agreed.

Buy and sell pricing periods rarely match, and the gap is a real position

The purchase prices over five days around the bill of lading at the load port. The sale to Helios may price over a different window, perhaps around the discharge date three weeks later. Those two windows do not overlap. Between them, the desk holds a genuine exposure to the benchmark moving, and it manages that with a time spread in futures rather than a flat position.

This is the single most common source of unintended exposure in physical trading. Two trades that each look hedged are, together, a calendar spread position that nobody explicitly chose. Whether the CTRM can see it depends entirely on whether both pricing periods were captured correctly at trade capture.

Quality adjustments settle the gap between specification and reality

The inspection results feed a quality adjustment clause. If sulphur, density or contaminants deviate from specification within permitted limits, the price adjusts by a formula. If they fall outside the limits, the buyer may reject the cargo or negotiate a settlement — which, with a vessel already at the berth, is almost always a negotiation rather than a rejection.

Figure 11 The margin waterfall: where $1.30 of gross differential goes

Step 8 — Getting paid is a documentary process, and the letter of credit is its central instrument

Commodities has its own version of delivery versus payment

In securities markets, delivery versus payment guarantees that the asset and the cash move simultaneously, so neither party can be left exposed. Physical commodities cannot do that: the cargo takes weeks to arrive and cannot be reversed. The market's substitute is documentary settlement — payment against documents rather than against goods.

Under a letter of credit, the buyer's bank irrevocably undertakes to pay the seller provided the seller presents documents conforming exactly to the credit's terms. The seller therefore takes credit risk on a bank rather than on the buyer. The documents typically required are the commercial invoice, the bill of lading, the inspection certificates for quantity and quality, the certificate of origin and the insurance certificate where applicable.

The banks deal in documents, not in goods. If the documents conform, the bank must pay even if the cargo is disputed. If they do not conform — a date wrong, a name spelled differently, a document missing — the bank may refuse, and the seller who has already shipped is exposed. Discrepancy rates on first presentation are high across trade finance generally, and resolving them is a real operational workload.

Open account and cash against documents cover the rest

Not every trade uses a letter of credit; they are expensive and consume credit lines. Established counterparties often trade on open account, paying an agreed number of days after the bill of lading, with the seller taking direct credit risk. Cash against documents sits between the two: documents are released through a bank only against payment, without the bank guaranteeing anything.

The choice among the three is a credit decision that belongs to the trade's economics. A letter of credit costs money and that cost comes out of the same margin computed above.

Provisional and final invoices reflect the pricing timeline

Because the price is not known at delivery, the seller typically issues a provisional invoice based on an estimated or provisional price shortly after loading, and a final invoice once actualisation completes. The difference is settled between the parties. Systems must therefore support two invoice events per cargo, track the difference, and prevent the common error of hedging or booking P&L against the provisional number as though it were final.

Settlement is where commodities meets the payments world, and the protocols change. A documentary credit is issued as a SWIFT MT700 and amended by MT707; guarantees travel as MT760. The payments themselves move as MT103 for customer transfers and MT202 for bank-to-bank, all migrating to the ISO 20022 pacs message family. This is an entirely different technology stack from the FIX messages that executed the futures hedge on the same trade — different systems, different teams, different latency expectations. Part 4 treats the split between these worlds as the concrete expression of the front-to-back divide.

Step 9 — The position lives on after the cargo is gone

Post-trade is not the end of the trade; it is most of its life

When the cargo discharges, several things remain open. The final invoice may not yet be settled. Demurrage claims and quality claims can stay unresolved for months. The futures hedge must be unwound or rolled as the position runs off. Regulatory reports must be filed and reconciled. And the P&L must be explained.

P&L attribution is where the business finds out what actually happened

Product control decomposes the result into its causes: how much came from the differential the trader actually traded, how much from the flat price the hedge was supposed to neutralise, how much from freight, from financing, from demurrage, from quality adjustments and from the timing mismatch between pricing windows.

This decomposition is the honest scorecard. A desk that made money because crude rallied while its hedge was imperfect did not earn that money; it was lucky, and the same imperfection will cost it next time. Attribution that cannot separate skill from residual basis exposure lets luck accumulate silently until it reverses.

Rolling and expiry management keep the hedge aligned

Futures contracts expire. A hedge for a cargo pricing next month must sit in the right contract month, and as the physical timeline shifts — a delayed vessel, an extended pricing period — the hedge must move with it. Rolling means closing the expiring contract and opening the next, and it costs the spread between them, which in a steeply backwardated market is expensive and in a steep contango is a benefit.

An expiring physically deliverable futures position that nobody rolled becomes a delivery obligation. Systems must alarm loudly on approaching expiry against physically deliverable contracts, and position limits near expiry exist precisely because the market's physical delivery capacity is finite.

Reconciliation is continuous and never finished

Positions must reconcile between the CTRM and the clearing broker, between the CTRM and the scheduling system, between invoices issued and cash received, and between what the bank reported to regulators and what it actually holds. Breaks are normal, and the discipline is in how quickly they are investigated rather than in whether they occur.

Part-cargo and term distribution reach the buyers a full cargo cannot

Not every buyer can take two million barrels

A VLCC parcel is too large for most refiners to purchase in one transaction, which creates a distribution problem. The market solves it three ways.

Part cargoes split a vessel among multiple buyers, with separate bills of lading for each parcel and separate inspection at each discharge. Co-loading combines several sellers' parcels into one vessel to achieve economic scale. Term contracts commit a buyer to a recurring smaller volume — a part cargo a month — which gives the seller predictable offtake and the buyer predictable supply.

Each adds operational complexity in exact proportion to the commercial benefit. Splitting one cargo among three buyers triples the documentation, the inspection events, the invoices and the potential for one buyer's delay to become the other two buyers' demurrage. Systems must model a single physical movement satisfying several contractual obligations, which is a genuinely harder data model than one trade equals one movement.

  1. Why does risk and effort peak twice in a commodity trade rather than once?

  2. Which single field on a physical deal ticket most often causes expensive errors, and why?

  3. What does a desk still own after it has fully hedged a cargo's flat price?

  4. Explain how a correctly hedged position can create a funding crisis.

  5. What is nomination, and why is it the commodities equivalent of allocation?

  6. Why does the bill of lading matter so much more than its name suggests?

  7. What is actualisation, and what would go wrong if a desk booked P&L against the provisional invoice?

  8. What is demurrage, and why does it matter disproportionately in commodities?

  9. What is a netback, and what does it tell the desk that the build-up cannot?

  10. A cargo worth Brent plus $1.20 is hedged with futures, and Brent falls ten dollars while the differential narrows to plus $0.

  11. What is the net result on two million barrels, and what does it represent?

  12. Where is the line between legitimate blending and misconduct?

  13. Why is a letter of credit described as the commodities equivalent of delivery versus payment?

2

  1. Why does risk peak twice in a commodity trade?

  2. Which deal-ticket field most often causes expensive errors?

  3. How can a correctly hedged position create a funding crisis?

  4. Which is NOT a form of basis risk after a flat-price hedge?

3Roles, Actors and Interactions

Part 2 followed the trade. This part takes the same trade apart by who does what. Five groups of external actors, one desk in the middle, six internal teams, and a wall running through all of them.

Five camps meet around every commodity trade, and the desk touches all of them

The camps have genuinely different objectives, which is why an intermediary exists

Producers and sellers want price certainty and reliable offtake. National oil companies, independent producers, miners, smelters and farmers have capital committed to production that cannot be switched off when prices fall. They sell forward to lock revenue, they hedge to protect project economics, and they value a counterparty who will take volume consistently rather than opportunistically.

Consumers and end-users want cost certainty and reliable supply. Refiners, utilities, airlines, chemical producers, food manufacturers and industrial fabricators have the mirror-image problem. Their input costs are volatile and their output prices frequently are not, and a hedge that fixes the input is the difference between a viable margin and a lost year.

Trading houses are the desk's competitors, counterparties and customers at the same time. Independent physical traders exist to arbitrage location, grade and time. They compete with the bank for the same cargoes, they trade with the bank constantly, and they are also major clients of the bank's financing and hedging business. That triple relationship is normal in commodities and requires careful conflict management.

Financial participants provide liquidity and take the other side. Hedge funds, commodity trading advisers, index investors and other banks trade the price without any intention of touching the product. They are frequently criticised as speculators, and that framing misses the point: when a producer sells a year of forward production, someone has to buy it. Financial participants are the natural counterparty to the hedging community's structural one-sidedness.

Market infrastructure and service providers make the whole thing enforceable. Exchanges and clearing houses, price reporting agencies, independent inspectors, shipowners and brokers, warehouse and terminal operators, and the banks that issue letters of credit. None of them takes a price view. All of them are indispensable, and several of them can stop a trade dead.

Figure 12 The desk sits at the centre: producers and consumers on the commercial axis, infrastructure and service providers underneath

Every stage of Part 2 is a specific interaction between two of these parties

From To Interaction Stage
Producer Desk Offers cargo, agrees differential and loading window 1-2
Desk Consumer Shows price, agrees sale on a delivered basis 1-2
Desk Exchange / CCP Sends hedge orders, receives fills, posts margin 4, 9
Desk Counterparty Exchanges and matches trade confirmations 3
Desk Shipbroker / owner Charters vessel, agrees freight rate and laycan 5
Buyer Seller Nominates vessel; seller nominates terminal and loading range 5
Desk Inspector Appoints inspection, receives quantity and quality certificates 6
Carrier Desk Issues bill of lading on completion of loading 6
PRA Whole market Publishes daily assessments that fix contractual prices 7
Desk Issuing bank Presents documents under letter of credit, receives payment 8
Desk Regulator / repository Reports derivative and wholesale energy transactions 9

The counterparty taxonomy decides how the desk prices, hedges and worries

Four types of counterparty, four completely different sets of behaviour

Understanding who is on the other side is not relationship management trivia. It determines the price the desk shows, the credit terms it offers, how it hedges the resulting position, and what it fears.

Producers hedge structurally and in one direction

A producer is a natural seller. It hedges by selling forward, and it does so year after year in the same direction, because its underlying exposure never flips. Its horizon is long — hedging programmes often run several years out — and its price sensitivity is moderate, because the objective is securing project economics rather than optimising a basis point.

The desk values producers because they generate large, predictable, long-dated flow. It fears them for one specific reason: when a producer's hedge moves against it, the producer owes collateral, and a producer whose hedge is deeply out of the money while its physical revenue has not yet arrived is exactly the counterparty most likely to have a liquidity problem.

Consumers hedge structurally in the opposite direction, and less consistently

An airline or a utility is a natural buyer. It hedges by buying forward, its horizon is typically shorter than a producer's, and — critically — its hedging discipline tends to be pro-cyclical. Consumers hedge enthusiastically after prices have risen and lose interest after they fall, which means consumer hedging demand arrives precisely when hedges are most expensive.

Consumers are prized because the flow naturally offsets producer flow: a desk that can internalise a producer's sale against a consumer's purchase earns the spread on both without carrying the risk. Matching those two is the core of the market-making franchise.

Trading houses are sophisticated, price-sensitive and fast

An independent physical trader knows exactly what the cargo is worth, often better than the bank does. It trades on thin differentials, moves in size, and will not pay for relationship. Flow from trading houses is informative — if a major house is aggressively bidding a grade, that is information — and it is also dangerous for exactly the same reason. Adverse selection is a real cost: the desk is most likely to be filled precisely when its price is wrong.

Financial participants trade the price and never the product

Hedge funds, CTAs and index investors bring liquidity, tighten spreads and absorb the hedging community's structural imbalance. Their horizon runs from intraday to a few months, their price sensitivity is extreme, and their positions can reverse quickly. Index-tracking flows have a distinctive feature worth knowing: they roll positions on a published schedule, which creates a predictable, concentrated pattern of activity in specific contract months.

Counterparty type Typical horizon Price sensitivity Why the desk wants them, and what it fears
Producers Multi-year, structural Moderate Wants large predictable one-way flow and financing business; fears collateral strain when hedges move against them
Consumers / end-users Months to a few years Moderate, pro-cyclical Wants flow that naturally offsets producers; fears demand disappearing exactly when the desk needs the other side
Trading houses Days to months Very high Wants volume, information and financing revenue; fears adverse selection and direct competition for the same cargoes
Financial participants Intraday to months Extreme Wants liquidity and the ability to lay off risk; fears sudden position reversals and crowded, correlated unwinds

Information barriers exist because one institution knows things it must not tell itself

The problem is structural, not a matter of trust

A bank running a commodities franchise accumulates information that would be enormously valuable to its own traders and that it is legally prohibited from using. This is not hypothetical.

The financing team knows a producer is negotiating a facility because a field is behind schedule. The corporate advisory team knows a utility is about to announce an acquisition that will change its fuel demand. The physical operations team knows a refinery has had an unplanned outage, because their cargo was rejected. Each of these is material non-public information, and each would be tradeable if the trading desk knew it.

Commodities has a second, less familiar category of inside information

In equities, inside information is about the issuer. In commodities, it is about physical fundamentals — an outage, a pipeline failure, an unannounced production cut — and the regimes that govern it are different. In the European Union, REMIT requires participants in wholesale energy markets to publish inside information about their own facilities and prohibits trading on it before publication. The Market Abuse Regulation extends to commodity derivatives and, in defined circumstances, to related spot contracts.

This creates an obligation that has no equity equivalent: a firm that owns physical assets may be required to disclose the very information its trading desk would profit from. Managing that is an information-barrier problem, a disclosure-workflow problem and a systems problem simultaneously.

Four functions must be separated, and the separation is enforced by technology

Private-side functions — commodity financing, structured transactions, corporate advisory — receive confidential client information and must not pass it to traders.

Public-side trading — market making and proprietary risk-taking on the desk — must operate only on public information.

Physical operations sit awkwardly between the two, because scheduling and logistics generate genuine fundamental information about outages, delays and demand that is commercially sensitive and sometimes legally restricted.

Research and market commentary, and any staff participating in PRA assessment windows, must be insulated from the desk's positions, because a published view that talks a position's book is manipulation.

Wall crossing is the controlled exception, and it is a logged event

When a private-side transaction genuinely requires trader expertise, the trader can be brought over the wall — formally, temporarily, with compliance approval and a written record. From that moment the trader is restricted from trading the affected instruments until the information becomes public or goes stale. Crossing is not a formality to be waved through; it is a decision that removes a revenue-generating individual from a market, and it is recorded because regulators will ask.

Information barriers are enforced by a stack of systems, not by a policy document. Access control segregates deal systems, document repositories and client data by side of the wall, so a trader cannot open a financing file. Restricted and watch lists, maintained centrally, block trading in affected names and instruments and are checked automatically at order entry. Wall-crossing logs record who was crossed, by whom, when, for what, and when they were released. Communication capture archives every channel — email, chat, recorded voice, and increasingly mobile and collaboration platforms — and feeds it into surveillance. Trade surveillance correlates communications with trading activity and flags patterns such as trading ahead of a client order, activity concentrated in an assessment window, or positions that move suspiciously before a public announcement. Personal account dealing systems police staff's own trading. None of these works alone, and a gap in any one of them makes the others decorative.

Six teams carry the trade, and the wall runs straight through the table

Each team owns a stage, a decision and a system

Team Division Core role Primary technology Side of the wall
Trading Markets, public side Prices risk, executes, owns the position and its P&L CTRM front end, pricing and curve engine, execution gateways Public
Structuring Markets, public side Designs tailored hedges and structured solutions Pricing libraries, scenario and payoff modelling tools Public
Sales and coverage Markets, public side Owns client relationships, distributes prices and axes CRM, quoting and RFQ platforms, client analytics Public
Operations and scheduling Markets operations Converts contracts into vessel movements; documents, inspection, demurrage Scheduling and logistics systems, vessel tracking, document management Operational
Middle office and risk Independent control Confirms trades, validates capture, measures and limits exposure CTRM risk modules, VaR and stress engines, limit monitoring Independent
Product control and finance Independent control Independent valuation, P&L attribution, invoicing, reconciliation Valuation control, P&L systems, invoicing and settlement platforms Independent

Two further functions sit across all six. Credit sets and monitors counterparty limits and collateral. Compliance owns surveillance, sanctions screening, restricted lists and regulatory reporting, and it is the only function with the authority to stop a trade for non-commercial reasons.

The reporting lines matter as much as the roles

Middle office, risk, product control and compliance do not report to the head of trading, and that is the entire point. The independence of valuation is a structural control: if the trader marks their own position, the P&L is an assertion rather than a measurement. Independent price verification — product control checking the desk's marks against observable market data — is the mechanism that catches both honest mismarking and deliberate concealment, and in commodities it is unusually demanding because so many positions reference illiquid differentials with no screen price.

Marking an illiquid book is where incentives quietly bend

Part 1 noted that a commodity desk's profitability is invisible on the ticket and only emerges once the position is valued. That fact has a consequence worth stating directly, because it is this market's equivalent of the league-table distortion familiar from capital markets.

Most of what a commodity desk holds has no screen price. A differential for a grade that trades twice a month, a forward curve five years out in a market liquid only for six, a location spread between two ports that rarely trade against each other, a structured hedge with an embedded option — none of these has an observable market level. Their value is produced by a model fed with judgement.

And the person best placed to produce that judgement is the trader, whose compensation depends on the answer. This is not an accusation; it is a structural conflict that exists in every trading business and is unusually acute here because the illiquid share of the book is unusually large.

Four distortions recur, and control functions look for all of them

Marking to the favourable edge. Where a range is defensible, the trader marks at the end of it that flatters the position. Each individual mark is arguable; the aggregate is a systematic bias.

Smoothing. Holding back unrealised gains in a good period to cover an expected weak one. It looks prudent and it is misreporting, because the P&L no longer corresponds to the period it is attributed to.

Reluctance to mark down illiquidity. When a position becomes hard to exit, its true exit value falls below its model value. Recognising that reduces today's P&L, so it tends to be recognised late.

Day-one profit on structured trades. A tailored transaction can be booked with an immediate gain representing the difference between the client price and the model value. If the model is optimistic, the desk has recognised profit that the market has not yet confirmed.

The defences are partly organisational and substantially technological

Independent price verification, introduced above, is the core control: product control sources its own market data, revalues the book independently, and investigates differences beyond a defined tolerance. Valuation adjustments then formalise what the model does not capture, typically a bid-offer reserve reflecting the cost of actually exiting, a liquidity reserve scaled to position size against market depth, and a model reserve where the valuation technique itself is uncertain.

None of this works without independent data. If product control revalues the book using curves sourced from the desk, it has verified nothing but arithmetic. Separate data sourcing is therefore an architectural requirement, not a procedural preference, and it is one of the clearest tests of whether a firm's control framework is real.

Three different truths: economics, cash and accounting

A single hedged cargo produces three different pictures of the same transaction, and confusing them causes genuine commercial errors.

The economic picture is what Part 2 described: the cargo and its hedge offset, the desk holds the basis, and the result is known at actualisation.

The cash picture is the funding gap: variation margin leaves daily while the physical gain arrives weeks later. Treasury owns this one.

The accounting picture is the one that reaches the reported earnings, and it can differ from both. Under prevailing accounting standards a derivative is carried at fair value with changes running through profit and loss, while a physical contract may be accounted for on an accrual basis or fall outside derivative accounting altogether under an own-use exemption. The hedge is therefore marked while the hedged item may not be, producing reported earnings volatility even when the economics are flat.

Hedge accounting is the formal election that allows the two to be matched, and it is not free. It requires designation and documentation at inception, a demonstrable economic relationship between hedge and hedged item, and ongoing effectiveness assessment. Fail any of those and the relationship is undone for accounting purposes even though the desk is still perfectly hedged in economic terms.

Mapping the six teams onto the nine stages

Trading and sales own stages 1 and 2. Middle office owns stage 3 and validates stage 2. Trading and risk share stage 4. Operations and scheduling own stages 5 and 6 outright, and they are the largest headcount in a physical business. Product control and middle office own stage 7. Operations, finance and trade finance share stage 8. Product control, risk and compliance own stage 9.

Read that distribution against the two peaks from Part 2 and the staffing model explains itself. The front office is dense at the first peak and almost absent at the second, where operations carries the trade alone.

  1. Why does a market with natural buyers and natural sellers still need financial participants?

  2. How does counterparty type change the price the desk shows for the same cargo?

  3. What is adverse selection, and which counterparty type creates most of it?

  4. How does inside information in commodities differ from inside information in equities?

  5. Why is physical operations described as sitting awkwardly relative to the wall?

  6. Name three systems that enforce information barriers and state what each contributes.

  7. Name two of the four recurring valuation distortions on an illiquid book, and the control that addresses them.

  8. Why is it insufficient for product control to revalue the book using the desk's own curves?

  9. How can a cargo be perfectly hedged economically and still create reported earnings volatility?

  10. Why must product control report outside the trading line?

3

  1. Why do commodity markets need financial participants?

  2. Which counterparty creates most adverse selection?

  3. How does inside information differ in commodities?

  4. Which flow is worth more to a desk that is long?

4Systems, Data and IT

The same trade again, this time as data. What functions any commodity business needs, who supplies them, how one deal ticket propagates through six systems, which protocols carry it, and why a platform that is merely correct is not sufficient.

Start from functions, not from vendors

Ask what any commodity trade needs before asking who sells it

Vendor names change and architectures outlive them. The durable question is what capabilities a commodity business requires regardless of who provides them, and the answer is six functional blocks in sequence plus four cross-cutting capabilities underneath.

Functional block What it must do Where it bites if it is weak
Pricing and pre-trade analytics Build forward curves for every grade, value structures, compute hedge ratios, show the trader a defensible price The desk quotes wrong prices and loses money it never sees
Trade capture (CTRM/ETRM) Represent physical and financial trades in full, link hedges, hold every contractual term Everything downstream inherits the error; unbookable products migrate to spreadsheets
Risk and position management Aggregate exposure by grade, location and tenor; compute VaR, Greeks and stress; enforce limits Exposure is discovered rather than managed
Logistics and scheduling Plan vessels and movements, manage nominations, track laytime and demurrage Cargoes miss windows; demurrage eats the margin
Confirmations and documents Match confirmations, manage contracts, GTCs, certificates and credit instruments Legal exposure accumulates silently in unconfirmed trades
Settlement, invoicing and operations Actualise prices, raise provisional and final invoices, present documents, reconcile cash Cash is late, claims go stale, working capital is destroyed

Underneath all six run market data and forward curves, compliance and sanctions screening, reference and static data, and regulatory reporting and workflow. These are not a seventh block; they are consumed by every block above.

Figure 13 Six functional blocks in sequence, four cross-cutting capabilities beneath them

The CTRM is the spine, and its scope is what makes commodities different

A Commodity Trading and Risk Management system is a single platform that holds trade capture, position and risk, and much of the physical and settlement lifecycle in one data model. Energy-specific implementations are usually called ETRM.

That combined scope has no real equivalent in other asset classes, where trade capture, risk and settlement are typically separate systems joined by interfaces. The reason is the physical leg. A cargo's position, its vessel, its inspection result and its invoice are the same object at different points in time, and splitting them across systems means reconciling the same trade against itself continuously. So the industry builds monoliths, accepts their weight, and spends the next decade trying to decompose them.

In practice there is rarely one CTRM, and that is the harder problem

The description above is the tidy version. The reality in almost every institution is several CTRMs, because the platforms are commodity-specific by design and no single product serves oil, gas and power, and metals equally well. A bank typically runs one platform for crude and products, another for gas and power with its own scheduling and nomination logic, and a third for metals and concentrates with warrant and assay handling the others do not need. Acquisitions add more.

This creates the problem that dominates commodity technology roadmaps: cross-commodity position aggregation. The bank's total exposure to, say, a rise in energy prices is spread across three systems with different data models, different curve conventions, different units and different end-of-day timings. Producing one consolidated risk number therefore requires a mapping layer, and that layer is where the errors live.

The consequence is that the aggregate risk number is a construction, not an observation. Someone decided how a gas position in therms and a crude position in barrels are expressed in common terms, which curve each is valued against, and what to do when the three systems struck their positions at different moments. Those decisions are usually documented somewhere and rarely challenged, and they are exactly what an incident investigation goes looking for.

The vendor landscape reveals where banks think their edge is

Buy the plumbing, build the judgement

The pattern across the industry is consistent and it is a map of competitive advantage.

Trade capture, scheduling, confirmations and settlement are bought. These are complex, well-understood, non-differentiating and expensive to build. The vendor market is concentrated, dominated by a small number of large platform providers offering broad CTRM and ETRM suites, alongside specialists focused on particular commodities or on lighter cloud-native deployments. No bank gains an advantage from a better demurrage calculator.

Market data is bought, and it is expensive. Exchange feeds, PRA assessments, freight rates and fundamentals data come from external providers under licences with strict redistribution terms. Those licensing constraints are a genuine architectural input: a system design that copies assessment data into six downstream databases can breach the licence, so entitlement management is a real engineering requirement rather than a procurement footnote.

Curve construction, pricing models and the risk engine's commodity-specific logic are built. How a bank builds a differential curve for a grade that trades twice a month, how it interpolates between illiquid points, how it models the optionality embedded in a tolerance clause — these are the judgements that separate a good desk from a mediocre one, and no bank outsources them.

Regulatory reporting is bought and heavily customised, because the rules are common but every firm's data model differs.

Optimisation is where commodity technology earns money rather than saving it

Most systems reduce cost; three of them create margin

Almost everything described so far exists to make the business correct, controlled and auditable. Three capabilities are different in kind, because they produce commercial value directly, and all three are optimisation problems.

Blend optimisation decides how to combine available components into a cargo that meets specification at the lowest cost, or that captures the most quality giveaway inside the tolerance. Part 2 introduced the commercial logic. As a computation it is a linear program: maximise blend value subject to every quality parameter — sulphur, density, vapour pressure, metals content — remaining within contractual limits, and subject to the physical availability of each component. Refiners have solved this for decades; trading desks apply the same technique to cargoes.

Scheduling and vessel optimisation decides which cargo goes on which vessel, in which sequence, to which port. The objective is to minimise freight and demurrage while honouring every laycan and nomination deadline. It is a routing and assignment problem with hard time windows, and the difference between a good schedule and a mediocre one is measured in the same cents per barrel that Part 2 showed the whole margin living in.

Nomination and allocation optimisation decides which physical parcel satisfies which contractual obligation when several of each exist. Because contracts differ in price, tolerance and quality requirement, and cargoes differ in cost, timing and assay, the matching is a genuine optimisation rather than a clerical assignment.

These three engines are the strongest argument for elastic compute in commodity technology, and they are usually the first workloads a desk moves to cloud. A blend or scheduling optimisation over a realistic component set is computationally heavy, the business wants to run it repeatedly with different assumptions, and the value of an answer decays quickly because the underlying availability changes hourly. A model that takes overnight to solve is a model the trader will not use; a model that solves in ninety seconds changes how the desk works. This is the one area where raw computational capacity converts directly into commercial margin rather than into faster reporting.

One deal ticket, six systems, and a cascade that cannot be stopped

Follow the object rather than the process

Figure 14 The deal ticket propagates through six systems, and each downstream object inherits every upstream error
Stage System Data object created What it depends on
Execution Trade capture (CTRM) Deal ticket: grade, volume, formula, delivery basis, counterparty, hedge link The trader's recap, correctly keyed
Risk Risk and position engine Position by grade, location and tenor; VaR; limit consumption The deal ticket and the curve set
Physical Scheduling and logistics Nomination, charter, laytime record, bill of lading, inspection certificate Delivery terms and dates on the ticket
Confirmation Confirmation platform Matched confirmation or executed contract Every commercial term on the ticket
Settlement Invoicing and settlement Provisional invoice, final invoice, credit instrument, payment The pricing formula, the B/L date and the inspection result
Reporting Regulatory reporting Transaction reports to trade repositories and regulators Counterparty identifiers and full trade economics

The cascade principle, stated plainly

Every downstream object is derived from an upstream one, so an error at the top reaches everything below it and gets harder to correct at each step. A pricing period keyed as the calendar month instead of five days around the bill of lading produces a wrong hedge in risk, a wrong provisional invoice in settlement, a wrong final price at actualisation, a wrong P&L attribution and a wrong regulatory report. Six systems are now confidently wrong in a mutually consistent way, which is far more dangerous than six systems disagreeing.

This is why controls concentrate at capture. Four-eyes checks on physical tickets, automated validation of pricing formulas against a library of permitted structures, mandatory hedge linkage before a physical trade can be released, and independent confirmation matching all exist to catch the error while it is still cheap.

The deal ticket is the atom, and its anatomy explains the whole architecture

Field by field, and what each one detonates downstream

Part 2 introduced the field groups. Here is why specific fields matter to the systems that read them.

Grade and quality specification drive which curve values the position. A cargo booked against the wrong grade is valued against the wrong differential curve, and the mismarking can persist for weeks because there is no screen price to contradict it.

Volume, unit and tolerance drive the hedge ratio and the invoice. Unit-of-measure errors are a classic and expensive failure: barrels and tonnes are related by a density conversion that varies by grade, and a system that hard-codes a single conversion factor will be wrong for every grade but one.

Unit conversion deserves to be a service, not a formula in each system

The barrels-and-tonnes problem is worth isolating because it is so reliably expensive. Crude volume in barrels converts to weight in tonnes through a density that varies by grade, and there is no single correct factor. Gas is quoted in therms, million British thermal units, cubic metres or megawatt-hours depending on the market, and conversion between energy and volume depends on the calorific value of the specific gas. Metals move between tonnes and pounds, and precious metals use troy ounces.

When each system carries its own conversion logic, they will disagree, and the disagreement surfaces as a reconciliation break that looks like a data error rather than a modelling error. Mature architectures expose conversion as a single shared service with versioned, auditable factors tied to the specific grade and, where relevant, to the specific cargo's assay. The bank should be able to answer the question "which density did we use for this cargo, and when did that value change?" — and a surprising number cannot.

The pricing formula — benchmark, differential, pricing period, basis, rounding — drives actualisation, hedge selection and P&L. It is the highest-consequence field group on the ticket and deserves the strongest validation.

Incoterm and delivery points drive risk transfer, freight cost, insurance and the point at which title moves. They also determine when the position stops being the desk's problem, which is a risk-system question as much as a legal one.

Counterparty identity, legal entity and netting agreement drive credit exposure, sanctions screening, netting sets and regulatory reporting. A counterparty booked against the wrong legal entity of the same group produces a correct-looking trade with wrong credit, wrong netting and a wrong regulatory report.

Hedge linkage drives whether the position is visible as hedged. This field has no commercial content and it is among the most important on the ticket.

Deduplication is a genuine engineering problem, not a data-quality chore

The same trade can arrive from a voice recap keyed manually, an electronic execution feed, a broker confirmation and a counterparty's confirmation submission. Naive matching creates two failure modes and both are expensive. Missing a duplicate creates a phantom position that consumes limits and may trigger an unnecessary hedge. Wrongly merging two genuine trades makes real exposure disappear.

Commodities makes this harder than most markets. Tolerances mean two legitimately distinct cargoes of the same grade in the same window can differ only in fields that are themselves approximate. Physical trades are amended frequently — dates shift, vessels change, volumes are revised at loading — so the record is a moving target rather than a fixed one. Matching therefore has to operate on a tolerance-aware, version-aware model, and it still needs human adjudication for the residual.

Identifiers are how systems agree they are talking about the same thing

Reference data is invisible until it is wrong, at which point it is the only thing anyone talks about.

Identifier What it names Where it bites
ISIN A specific listed security or standardised contract Identifies exchange-traded futures and options in reporting and clearing; does not exist for a physical cargo
LEI The legal entity that is party to a trade Mandatory in derivative and transaction reporting; wrong LEI means a rejected or misattributed report
UTI / UPI The unique transaction and product identifiers under derivative reporting regimes Both sides of a trade must report the same UTI, so mismatches create reconciliation breaks with the regulator
MIC The venue or market where execution took place Determines which reporting obligations and position limits apply
Grade and product codes The specific commodity grade and its assessment reference Selects the valuation curve and the settlement price; usually a mix of vendor codes and internal reference data
IMO number The vessel, permanently and uniquely Sanctions and vetting screening, tracking, demurrage and claims

The last two have no equivalent in other asset classes and both are load-bearing. A commodities reference data function must maintain a grade master describing quality parameters and their relationships to benchmarks, and a vessel and counterparty master feeding compliance screening. Neither is glamorous and both stop trading when they fail.

Three integration patterns, and the migration between them is not just about speed

Batch is the legacy default. Systems exchange files on a schedule, typically overnight. Positions are struck at end of day, risk is computed, reports are produced. It is simple and robust, and its weakness is latency: for the hours between batches, the risk system's view of the world is stale. In a market that can move twenty per cent in a session, an overnight risk number is a historical document.

APIs made data available on demand. A system requests the current position, the latest curve or a counterparty's credit availability and receives an answer immediately. This is the modernisation most commodity platforms have been through, and it is what makes intraday risk and real-time pre-trade credit checks possible.

Event streaming pushes changes as they happen. Every trade booked, every amendment, every fill, every nomination change and every market data tick is published to a durable log that any number of consumers subscribe to. Risk, limits, compliance surveillance and reporting all read the same stream and stay current without polling anything.

The move from batch to streaming is not a speed upgrade; it removes the need for reconciliation. When risk, settlement and reporting each pull their own overnight extract and transform it independently, they will disagree, and an entire function exists to explain the differences. When all three consume the same ordered event stream, they cannot disagree about what happened — only about how to interpret it. The reconciliation team does not disappear, but its work shifts from explaining differences that architecture created to investigating genuine breaks with external parties. That is the real prize, and it is why streaming migrations are justified by control improvement rather than latency.

Two governance problems sit underneath the integration pattern

A golden source is the single system designated as authoritative for a given data domain, and the discipline is that every other system consumes it rather than maintaining its own version. Counterparty static data, grade and assay reference data, curve marks and vessel data each need one. The failure pattern is familiar: two systems each hold a counterparty record, both are edited, neither is wrong locally, and the credit exposure aggregates to the wrong legal entity.

End-user computing — spreadsheets, in plain language — is the other governance problem, and it is larger in commodities than almost anywhere else. Every product the CTRM cannot represent ends up in a spreadsheet next to it, and physical trading generates such products constantly: a novel structure, an unusual quality clause, a new market the platform has not been configured for. The spreadsheet then holds real exposure that the official systems cannot see.

The correct response is not to ban them, which fails, but to govern them: an inventory of which spreadsheets carry material exposure, version and access control on those, independent validation of their calculations, and a standing programme to migrate the material ones into the platform. Regulators examine this directly, and an unregistered spreadsheet holding a live position is treated as a control failure rather than as pragmatism.

Two things are always proprietary, and both for good reasons

The curve engine is the desk's intellectual property. Constructing a forward curve for a grade that trades sporadically requires interpolation, blending of observable and assessed prices, and judgement encoded as method. Two banks with identical market data will produce different curves, value the same position differently and quote different prices. This is the model risk that matters most in commodities and it is governed accordingly, with independent validation and periodic back-testing against realised prices.

The audit trail must be immutable and append-only. Every trade version, every amendment, every limit override, every wall-crossing, every mark and every communication is written once and never modified. Corrections are new records referencing old ones. This is not a preference; regulators, auditors and litigation all require the ability to reconstruct exactly what the firm knew and did at a point in time. A system that allows a record to be edited in place cannot answer that question, and in a dispute the inability to answer is treated as the answer.

Three protocols carry one trade, and they belong to different worlds

Figure 15 FIX executes, FpML confirms, SWIFT pays: three standards, three timescales, one trade
Layer Standard What it carries Timescale
Execution FIX Orders, cancels, fills, allocations to and from exchanges and electronic venues Milliseconds
OTC confirmation FpML Full economic terms of swaps and options for matching and reporting Minutes to hours
Physical documentation Contracts, GTCs, certificates, increasingly electronic bills of lading Title, quality, quantity, delivery obligations Days to weeks
Payment and credit SWIFT MT, migrating to ISO 20022 Documentary credits (MT700), amendments (MT707), guarantees (MT760), payments (MT103, MT202, pacs family) Hours to days

FIX is the language of execution. It is a terse, session-based message protocol optimised for speed, and it carries the futures hedge from the desk's order management system to the exchange and the fill back again. It says nothing about cargoes, quality or delivery; it was never meant to.

FpML is the language of derivative terms. Where FIX describes an order, FpML describes a contract in full economic detail, which is what confirmation matching and transaction reporting require.

SWIFT is the language of money and documentary credit. It is slow by front-office standards and utterly unforgiving about format, and it is migrating to the richer, structured ISO 20022 message set — a migration that matters here because the extra structured data improves sanctions screening and reconciliation.

Physical documentation remains the weakest link. Bills of lading, certificates of origin and inspection certificates are still substantially paper or PDF. Electronic bill of lading platforms exist and legal frameworks have advanced — model law on electronic transferable records and national legislation giving electronic trade documents the same legal effect as paper — but adoption requires every party in the chain, including carriers, banks and terminals, to accept the same platform. That coordination problem, not the technology, is what has kept the paper original travelling by courier.

Resilience is where mission-critical becomes measurable

Some moments in a commodity trade have no second chance

Most systems can tolerate an outage by catching up afterwards. A few moments in this business cannot. A vessel's loading window is fixed and the berth is allocated; missing it means demurrage or a lost cargo. An exchange margin call has a deadline measured in hours, and failing it is a default event regardless of the reason. A PRA assessment window is a few minutes long and does not reopen. A letter of credit has an expiry date, and documents presented late are documents unpaid.

Cyber risk reaches the physical world through the billing system

Commodity businesses sit on an unusual boundary. Trading platforms are information technology, but they connect to operational technology: terminal automation, pipeline control, metering, tank gauging and grid dispatch. The two worlds have different patch cycles, different lifespans and, historically, different security assumptions.

The instructive real case is the ransomware attack on a major United States refined products pipeline in May 2021. The malware affected the company's business systems rather than the pipeline's control systems, yet the operator halted physical pipeline flows anyway, because it could not reliably measure and bill for what it was shipping. Fuel supply along much of the eastern seaboard was disrupted for days.

Read that carefully, because it inverts the intuition. The attack never touched the machinery. It touched the system that turns a physical movement into a commercial record, and without that record the physical movement could not responsibly continue. In a business where every molecule that moves must be measured, allocated, invoiced and reported, the commercial systems are not downstream of operations. They are a precondition for operations.

The design implications follow directly: segment IT from OT rather than assuming a shared perimeter, ensure the manual fallback procedures in the table below actually work when the primary systems are assumed hostile rather than merely unavailable, and treat backups as a recovery capability only if they have been tested against a scenario in which the attacker had access before the backup was taken.

Concept What it guarantees Why it matters on the critical day
Redundancy No single component failure stops the service A failed server does not stop a margin payment, but redundancy alone does not survive a site or region loss
Automatic failover Traffic moves to standby capacity without human intervention Manual failover during a volatility spike takes longer than the margin deadline allows
Disaster recovery The service can be restored at another site within a defined objective Recovery time and recovery point objectives define how much time and data can be lost — both are commercial decisions, not technical ones
Business continuity The business function continues even if the system does not Documented manual procedures for nomination and payment instruction when the platform is unavailable
Rehearsal The plan actually works An untested DR plan is a document, not a capability; regulators now examine evidence of testing, not existence of plans

Five regulatory regimes, read as system specifications

Each of these is usually presented as law. Read them instead as requirements documents, because that is what they become once they reach the architecture.

Derivative reporting and clearing rules define the reporting pipeline

The post-crisis reforms — EMIR in the European Union, the Dodd-Frank Act in the United States, and their equivalents elsewhere — require standardised OTC derivatives to be centrally cleared, margined if uncleared, and reported to trade repositories. As a specification this means: assign a unique transaction identifier agreed with the counterparty, capture the full product taxonomy, report within a deadline measured in one business day, reconcile continuously against the repository, and compute and exchange margin daily on uncleared positions. Every one of those is a system requirement with a hard deadline attached.

Position limits and reporting constrain what the desk may hold

Commodity derivative position limits, established under MiFID II in Europe and by the Commodity Futures Trading Commission in the United States, cap the net position any participant may hold in specified contracts, with exemptions available for genuine hedging. As a specification: aggregate positions across all entities and accounts in real time, map every position to the correct limit, distinguish hedging from speculative positions and evidence the distinction, and alert before a breach rather than after it. Aggregation across group entities is the hard part, and it is a data problem rather than a legal one.

Wholesale energy market integrity rules govern information and behaviour

REMIT in the European Union prohibits market manipulation and insider trading in wholesale energy markets and obliges participants to publish inside information about their own assets. As a specification: capture and report wholesale energy transactions and orders, maintain a disclosure workflow for asset-related inside information with timestamps, and run surveillance capable of detecting manipulation patterns specific to energy markets.

Market abuse rules extend to commodity derivatives and benchmark windows

The Market Abuse Regulation and its equivalents cover commodity derivatives and, in defined circumstances, related spot contracts. Benchmark regulation adds governance requirements around contributing to and administering indices. As a specification: capture every communication channel, retain records for the mandated period, run surveillance correlating communications with trading, apply enhanced controls and pre-approval to assessment-window activity, and maintain restricted lists enforced at order entry.

Sanctions and trade compliance can stop a trade at any point in its life

Sanctions regimes administered by bodies including the United States Office of Foreign Assets Control, the European Union and the United Kingdom prohibit dealings with designated parties and, in commodities, with specific cargo origins, vessels and routes. As a specification: screen counterparties, vessels, owners, flags, origins and routes at onboarding and continuously thereafter, re-screen automatically on every nomination or itinerary change, integrate vessel tracking to detect transponder gaps and undisclosed transfers, and block settlement where screening fails. This is the only regime on this list that can require a firm to stop a cargo already at sea.

  1. Why does the CTRM have a broader scope than the equivalent systems in other asset classes?

  2. What does the buy-versus-build boundary tell you about a bank's strategy?

  3. State the cascade principle and give a concrete example. — Every downstream object derives from an upstream one, so an error at capture reaches everything below it. A pricing period keyed as a calendar month instead of five days around the bill of lading produces a wrong hedge, wrong invoices, a wrong final price, wrong P&L attribution and a wrong regulatory report.

  4. Why is deduplication harder in commodities than in most markets?

  5. What do FIX, FpML and SWIFT each carry, and over what timescale?

  6. Why is the migration from batch to event streaming justified by control rather than by speed?

  7. Why is redundancy insufficient for operational resilience?

  8. Why is a bank's consolidated commodity risk number a construction rather than an observation?

  9. Name the three optimisation capabilities that create margin rather than control, and say what each solves.

  10. What did the 2021 pipeline ransomware incident demonstrate about the IT-OT boundary?

  11. Why must sanctions screening in commodities be continuous rather than point-in-time?

4

  1. Why is the CTRM's scope broader than equivalent systems elsewhere?

  2. What does the buy-versus-build boundary reveal?

  3. Why is deduplication harder in commodities?

  4. A wrong pricing period is captured. Where does the error show up?

5Risks, Failures and Transformation

What the desk can lose, how to tell whether it earned its result or was handed it, three occasions when the industry destroyed itself in three different ways, and which of today's technology changes are real.

Six risk families cover everything a commodity desk carries

The taxonomy is a map of controllability, not a list

Figure 16 Six risk families, grouped by who owns them and how they are defended
Risk family What it is Who owns it Main defence
Market / flat price The value of an unhedged position moves with the benchmark Trading, within limits set by risk Hedging with futures and swaps; VaR and stress limits
Basis / execution The hedge and the exposure do not move together: grade, time, location, volume tolerance Trading Explicit basis limits, curve validation, hedge-linkage discipline
Credit / counterparty The counterparty fails to pay, to deliver or to lift Credit Limits on potential future exposure, letters of credit, netting, collateral
Physical / delivery The cargo is late, off-spec, short, damaged, or cannot be discharged Operations Inspection, tolerances, laytime terms, vetting, insurance
Liquidity / funding Margin and working capital demands exceed available cash even when the position is economically sound Treasury, with risk Committed facilities, collateral forecasting, liquidity stress testing
Compliance / conduct Sanctions breach, market manipulation, misuse of inside information, reporting failure Compliance, with the business Screening, surveillance, restricted lists, controls on assessment-window activity

Reading the table properly is the point of it

The first two families are the risks the desk is paid to take. It exists to price and warehouse them, and eliminating them entirely would eliminate the business.

The next two are risks it must manage down to as close to zero as is economically sensible. Nobody is paid for demurrage or for a counterparty default; these are frictions to be minimised through process, documentation and insurance.

The last two are risks that must be prevented outright, because their tails are not proportional. A funding failure or a sanctions breach does not produce a bad quarter. It produces a solvency event or a regulatory action that can end the franchise, which is why both sit outside the trading line entirely.

The same six families wear different clothes in each market

The taxonomy is stable across commodities; what changes is which family dominates and how it presents. A risk framework written only for crude will systematically under-police the others.

Risk family In crude and products In base metals In power In agriculture
Market / flat price Benchmark moves on a large unhedged cargo Exchange price moves on financed inventory Extreme hourly spikes and negative prices Harvest and weather repricing the whole curve
Basis / execution Grade, location and pricing-window mismatch Physical premium diverging from the exchange price Shape and imbalance risk: the profile delivered differs from the profile hedged Origin and quality differentials, old crop against new
Credit / counterparty Buyer fails to pay against documents Borrower defaults on inventory financing Retail supplier or generator insolvency Farmer or cooperative fails to deliver after a bad harvest
Physical / delivery Late vessel, off-spec cargo, demurrage Warehouse queue delaying load-out; title fraud Plant outage or transmission constraint at the delivery hour Spoilage, moisture, infestation, port congestion
Liquidity / funding Margin on hedges while the cargo is unpaid Financing the stock through the queue Collateral demands during a price spike, which can be sudden and enormous Seasonal working capital peaks at harvest
Compliance / conduct Sanctions on cargo, vessel or origin; assessment-window conduct Warehouse and warrant integrity Manipulation of a market with a physical bottleneck, under REMIT Export bans and licensing regimes changing overnight
Two risks in this table have no crude equivalent and deserve naming

Shape and imbalance risk in power is the gap between the profile a desk hedged and the profile it actually delivered. A hedge bought as a flat block across every hour does not cover a customer whose demand peaks on winter evenings, and the difference is settled at balancing prices, which are precisely the most volatile prices in the market. A power book can be fully hedged in volume and badly exposed in shape.

Sudden collateral escalation appears in every market but is most violent in power and gas, because those markets can reprice by a multiple rather than by a percentage. The 2022 European energy crisis showed that the binding constraint on a hedged utility is not solvency but the speed at which it can post cash, and that the counterparty most likely to fail is one whose hedge is working exactly as designed.

Telling a good trade from a lucky one requires measuring the right things

Profit is not evidence of skill until it has been decomposed

A commodity desk can make money for reasons that have nothing to do with its judgement. The benchmark moved in its favour while its hedge was imperfect. A vessel arrived early and demurrage was avoided by chance. A counterparty accepted an off-spec cargo it could have rejected. None of these is repeatable, and a business that cannot tell them apart from skill will scale exactly the wrong activities.

Metric What it measures Why it matters
P&L attribution quality How much of the result is explained by identified factors rather than residual A large unexplained residual means the desk does not know why it made money and cannot repeat it
Basis capture versus flat-price P&L Whether profit came from the differential the desk chose or from the flat price it claimed to hedge Flat-price P&L on a supposedly hedged book is a control failure, not a win
Risk-adjusted return on limit used Profit relative to the risk capacity consumed Two desks with identical P&L are not equally good if one used three times the limit
Operational loss ratio Demurrage, quality claims, discrepancy costs and settlement errors as a share of gross margin On sub-dollar margins this is the difference between profit and loss, and it is entirely controllable
Confirmation and reconciliation ageing How long trades sit unconfirmed and breaks sit unresolved The most reliable early warning of an operational problem, and a leading indicator regulators watch
Cash conversion cycle Days between paying for a cargo and being paid for it Directly drives the financing cost that Part 2 showed consuming half the margin

Three real failures, three distinct ways to lose

These are not random cautionary tales; they are the three failure modes of the process

Metallgesellschaft, 1993: the hedge was right and the funding was wrong

The German industrial group's US energy subsidiary sold long-dated forward contracts to supply oil products to customers at fixed prices over up to ten years. To hedge, it bought short-dated futures and rolled them forward continuously — a stack-and-roll hedge, stacking the entire long-term exposure into near-month contracts.

The economic logic was defensible. The cash flow logic was not. When oil prices fell sharply and the market moved into contango, two things happened at once. The futures position lost money daily and required cash margin immediately. The offsetting gains on the forward supply contracts were unrealised and would arrive over the following decade. Rolling the stack in contango also cost money at every roll.

The firm faced margin demands it could not fund, the positions were liquidated near the bottom, and losses reached roughly 1.3 billion dollars. A hedge that might have worked over ten years failed over ten months because nobody had modelled the cash.

Failure mode: liquidity and funding. It maps to Step 4 of the lifecycle, hedging and risk transfer, and specifically to the margin asymmetry described there.

Amaranth Advisors, 2006: concentrated basis risk in an illiquid spread

The hedge fund built enormous positions in natural gas calendar spreads — long later-dated months against short nearer ones — expressing a view on the shape of the curve rather than its level. The positions were not merely large; they represented a dominant share of the open interest in the specific contracts involved — open interest being the total number of contracts outstanding and not yet closed, which is the standard measure of how much genuine capacity a contract has to absorb a position.

When the spread moved against the fund, the position could not be exited. Every attempt to sell moved the price further against it, because the fund was effectively the market. Losses of approximately 6.6 billion dollars accumulated in weeks and the fund collapsed.

The instructive detail is that this was not a naked directional bet on gas prices. It was a spread, which is to say a basis position — precisely the kind of exposure the desk in Part 2 deliberately retains after hedging. Basis positions are smaller per unit than flat price, which makes them feel safer, and they are far less liquid, which makes them far harder to exit.

Failure mode: basis risk plus concentration and illiquidity. It maps to Step 4, and to the basis exposures that survive every hedge.

Sumitomo, 1996: control failure over a decade

A single copper trader at the Japanese trading house accumulated unauthorised positions and concealed losses over roughly a decade, while attempting to control the physical copper market to support prices. When the scheme unravelled, losses were approximately 2.6 billion dollars.

The failure was not analytical. It was structural: inadequate segregation between trading and settlement, insufficient independent verification of positions and marks, excessive authority concentrated in one individual, and a control function without the standing to challenge a highly profitable trader. Every control described in Part 3 — independent product control, four-eyes on capture, independent price verification, surveillance — exists in its modern form partly because of cases like this one.

Failure mode: conduct and control. It maps to Steps 2 and 9, trade capture and independent verification.

Case Failure mode Lifecycle step it maps to The lesson that stuck
Metallgesellschaft, 1993 Liquidity and funding mismatch Step 4 — hedging and margin An economically sound hedge can be fatal if its cash timing is not modelled
Amaranth, 2006 Basis concentration in an illiquid market Step 4 — residual basis risk Position size must be measured against market liquidity, not just against limits
Sumitomo, 1996 Conduct and control failure Steps 2 and 9 — capture and verification Independent valuation and segregation are structural necessities, not overhead

Five technology changes are genuinely reshaping the business

Trend Attaches to Why it matters
Advanced analytics and machine learning in forecasting Step 1 — price discovery Satellite imagery of storage, vessel tracking, weather and crop models, and load and renewable output forecasts turn fundamentals into near-real-time signals, moving the information edge from access to interpretation
Electronic execution and algorithmic hedging Steps 2 and 4 Automates the liquid hedging leg, letting scarce trader attention concentrate on differentials and structure where the margin actually is
RegTech automation Steps 3, 8 and 9 Automated reporting, continuous sanctions and vessel screening, and surveillance across every communication channel
Cloud and API-first CTRM architecture All stages Breaks the monolith into services, enables intraday risk and elastic capacity for stress runs that used to take a weekend
Straight-through processing and digital documentation Steps 3, 6 and 8 Electronic bills of lading and digital confirmations attack the documentary layer, the slowest and most manual seam in the entire lifecycle

Apply one filter to every item on that list: does it change the economics or the risk, or does it change the appearance? Electronic bills of lading change economics, because they compress the cash conversion cycle that Part 2 showed consuming roughly half the trade's margin, and they remove the letter-of-indemnity exposure created when paper travels slower than the ship. Intraday risk changes risk, because a desk that knows its exposure at eleven in the morning can act on it and one that learns it overnight cannot. A redesigned trader dashboard changes the appearance. A blockchain pilot that requires every carrier, terminal, inspector and bank in the chain to join before it delivers value changes nothing until the last of them does — which is why the digital trade documents story has been about legal frameworks and network adoption rather than about the technology, which worked years ago.

Generative AI has a genuine place, and a narrow one

Where large language models actually help

Contract and documentation review is the strongest use case. Physical commodity trading generates enormous volumes of unstructured text: general terms and conditions, charter parties, letters of credit, inspection certificates, claim correspondence. Extracting structured terms from a contract, comparing a counterparty's GTCs against the bank's standard, and flagging deviations in sanctions or force majeure clauses are tasks where the technology is strong and the alternative is a person reading slowly.

Operational exception handling is the highest-volume use case. Letter-of-credit discrepancies, demurrage claim correspondence and confirmation breaks are repetitive, text-heavy and rule-bound. Drafting the first response, summarising a claim file and proposing a resolution based on precedent are all genuinely useful, with a human approving the outcome.

Research synthesis and market commentary is the most visible use case. Digesting fundamentals reports, inventory statistics, outage notices and news into a coherent summary for the desk saves real time, provided the output is treated as a starting point rather than a conclusion.

Counterparty and vessel due diligence is the fastest-growing use case. Assembling and summarising open-source information on a counterparty, its ownership chain or a vessel's history supports the compliance screening described in Part 4, again as a first pass reviewed by an analyst.

Use of GenAI The gain The governing risk
Contract and GTC review Faster review, consistent deviation detection Confidentiality of counterparty terms; a missed clause is a legal exposure
Operational exception handling Materially lower cost per claim and per discrepancy Errors settle real cash; every output needs human approval
Research and commentary synthesis Faster situational awareness for the desk Hallucinated figures entering a trading decision or a client document
Counterparty and vessel due diligence Broader coverage of open-source screening False negatives create sanctions exposure; the tool supplements, never replaces, screening systems

Three risks govern all of it

Confidentiality and inside information. A commodities franchise holds client hedging positions, physical fundamentals that may constitute inside information under REMIT, and private-side financing data. Any tool that could carry that information across an information barrier — or outside the firm — is an information-barrier breach in a new wrapper. The barriers in Part 3 must extend to model access, prompts and retrieval scope.

Hallucination and accuracy. A model that invents a plausible differential, a wrong density conversion or a non-existent contract clause produces an error that looks exactly like a correct answer. In a business where a unit-of-measure mistake misprices a cargo by millions, plausible-looking wrongness is the specific danger.

Explainability and model risk governance. Any model influencing pricing, risk or a regulatory outcome falls under model risk governance: documented purpose, validated performance, defined limitations, monitored drift and a named owner. A model that cannot explain why it produced an output cannot be used where the firm must justify a decision to a regulator, a client or a court.

  1. Why are the six risk families grouped rather than merely listed?

  2. What is shape risk in power, and why can a book be fully hedged and still badly exposed?

  3. Why did the 2022 European energy crisis threaten utilities whose hedges were working correctly?

  4. Why is a large profit with a large unexplained attribution residual a warning sign?

  5. What exactly went wrong at Metallgesellschaft, and which lifecycle step does it map to?

  6. Why is the Amaranth case a lesson about basis rather than about direction?

  7. Apply the transformation filter to electronic bills of lading. — They change economics. They compress the cash conversion cycle that consumes roughly half the margin on a physical trade, and they remove the contingent exposure created by letters of indemnity issued when paper documents travel slower than the vessel.

  8. What are the three governing risks of using GenAI on a commodities desk?

  9. Why is the human in the loop described as a control rather than a limitation?

5

  1. What is shape risk in power?

  2. Why did the 2022 energy crisis threaten utilities with working hedges?

  3. Why is a big profit with a big unexplained residual a warning?

  4. Why group the six risk families rather than list them?

Module Summary — The Six Ideas That Matter Most

If you remember nothing else from this module, remember these six. Each one has appeared in every part, and together they are the argument.

Idea 1 — The physical leg defines the platform, not the notional

A billion-dollar swap book needs a pricing engine, a risk system and a clearing connection. A two-hundred-million-dollar cargo needs all of that plus scheduling, chartering, inspection, documentation, letters of credit and demurrage management. Complexity in commodities scales with the number of physical decisions, not with money at risk. This is why the CTRM is a monolith, why operations is the largest team in a physical business, and why a firm's product mix and its architecture are the same decision made twice.

Idea 2 — Data cascades, so errors are never contained

The deal ticket is the atom. More than a hundred fields, keyed in the minutes after a trader says yes, determine the position, the hedge, the schedule, the invoice, the P&L and the regulatory report. A wrong pricing period or an unlinked hedge does not stay in the CTRM; it reaches six systems, which then agree with each other confidently and wrongly. Nothing downstream can be more accurate than the atom, which is why control investment belongs at capture rather than in downstream analytics.

Idea 3 — The price is fixed after the trade, not at it

This single fact makes physical commodity trading strange. The desk commits to a cargo weeks before its price exists, because the price is a formula applied to published assessments over a window that has not yet happened. Everything follows: the hedge must be placed against the pricing period rather than the trade date, buy and sell windows that do not overlap create a calendar position nobody chose, and a provisional invoice is not a result. Actualisation, not execution, is when the desk finds out what it did. The exception proves the rule — in power, a day-ahead auction fixes the price for every hour before delivery, and it is precisely because settlement is immediate there that the risk migrates instead into shape and imbalance.

Idea 4 — Compliance is core architecture, not an afterthought

Sanctions can stop a cargo already at sea. Position limits cap what the desk may hold and must be aggregated across every entity in real time. Reporting deadlines are measured in one business day and require identifiers agreed with the counterparty. Assessment-window activity needs pre-approval and surveillance. None of this can be bolted onto a finished platform, because each requirement reaches into trade capture, scheduling and settlement simultaneously. Screening that runs only at onboarding is not a control, and the firm that discovers this does so when a bank refuses to pay against its documents.

Idea 5 — Basis and funding quietly distort behaviour

The two risks that survive a good hedge are the two that have caused the largest losses. Basis feels safe because it is smaller per unit, and it is illiquid, so a position that cannot be exited is precisely the one that felt conservative when it was built. Funding feels irrelevant because the position is economically hedged, and cash does not care about economics: variation margin is paid today while the offsetting physical gain arrives in two months.

A third distortion compounds both. Because most of the book has no screen price, its value is a judgement produced by the person paid on the answer, as Part 3 described. Illiquidity therefore makes a position simultaneously harder to exit, harder to fund and easier to mark favourably — and those three properties travel together. The positions a desk should be most sceptical about are exactly the ones its own reporting is least equipped to challenge, which is why independent valuation with independently sourced data is a risk control rather than an accounting formality.

Idea 6 — Real transformation moves cost or risk

The changes worth funding compress the cash conversion cycle, move risk from overnight to intraday, remove reconciliation by removing the divergent copies that create it, or replace a manual control with a better one. Generative AI belongs on that list where it reads contracts, handles exceptions and supports due diligence, governed against confidentiality, hallucination and model risk. It does not belong where the decisive variables are physical, sparse and absent from any training set. The filter is the same for every technology and it has not changed: economics, risk, or appearance.

Process Cheat Sheet — The Whole Trade on One Page

A single reference for the entire commodity trade flow, from the price the desk shows to the cash it finally receives.

Stage Timing Core activity Key output Lead and systems
1. Price discovery Continuous Build benchmark + differential + quality + freight; respond to RFQ or show axes A defensible price Trading; curve engine, market data, PRA feeds
2. Execution and capture Minutes Agree terms by voice, chat or screen; key the deal ticket Booked ticket with hedge linkage Trading and middle office; CTRM
3. Confirmation T+0 to T+1 Match confirmations; agree contract and GTCs Enforceable contract Middle office and legal; confirmation platform, FpML
4. Hedging Same day Sell or buy futures against the physical; post initial margin Linked financial leg, flat price neutralised Trading and risk; execution gateway, FIX, CCP
5. Scheduling Days to weeks Nominate vessel and terminal; charter; vet vessel Agreed loading window and fixed freight Operations and chartering; scheduling, AIS, vetting
6. Delivery The loading and discharge windows Load, inspect, issue documents; track laytime Bill of lading, inspection certificates Operations and inspectors; logistics, document management
7. Actualisation The pricing period Apply the formula to published assessments; adjust for quality Final price per unit Middle office and product control; CTRM, PRA data
8. Invoicing and settlement 30 to 60 days from B/L Issue provisional then final invoice; present documents Cash received Operations, finance, trade finance; SWIFT MT/MX
9. Post-trade To expiry and beyond Roll or unwind hedge; settle claims; attribute P&L; report Reconciled position and filed reports Product control, risk, compliance; reporting stack
Metric Rule of thumb (illustrative)
Full lifecycle, physical cargo 2 to 3 months from agreement to final settlement
Time to agree a cargo Minutes, by voice or chat
VLCC parcel size ~2.0 million barrels; Suezmax ~1.0m; Aframax ~0.5-0.8m
Typical physical gross differential margin Cents to low dollars per barrel; often well under 1% of notional
Financing cost on a 45-day cargo Roughly $0.60 per barrel at mid-single-digit rates on ~$84/bbl
Demurrage on a VLCC Tens of thousands of dollars per day beyond laytime
Volume tolerance on a physical contract Commonly +/- 5%, and the holder of the option matters
Hedge ratio, crude cargo to futures ~1,000 barrels per ICE Brent or NYMEX WTI contract
Derivative reporting deadline Typically one business day after execution
Pricing period, crude cargo Commonly a 5-day window around the bill of lading date
Payment terms Commonly 30 days from B/L, against documents or letter of credit
Net margin after costs, worked example ~$0.52 per barrel on a $1.30 gross differential
Refinery netback ceiling, worked example ~$88.40 per barrel — the most a rational refiner could pay
Hedge residual on a 25-cent differential move ~$500,000 on a 2m bbl cargo — the risk the desk chose
LME liquidity benchmark The 3-month prompt date, rolling forward every day
Family Price reference How title and delivery work The main departure from the crude case
Crude and products Assessed benchmark plus differential, fixed after the trade Cargo, bill of lading, vessel This is the reference case worked through Part 2
Base metals Exchange settlement, daily prompt dates to three months Warrant against approved warehouse stock Title is a registry entry; queues create a physical premium
Power Sequence of forward, day-ahead, intraday and balancing prices Scheduled delivery to a grid node No storage, so no carry; prices can go negative or spike for one hour
Gas and LNG Hub price, or assessed marker for LNG Pipeline nomination, or LNG cargo Seasonality drives storage value; destination flexibility is a priced option
Agriculture Exchange futures plus origin and quality differentials Bulk vessel, elevator, warehouse receipt Annual harvest splits the curve into old and new crop

The whole trade in one sentence: the desk agrees a cargo at a differential to a benchmark whose level it immediately hedges away, ships it under a contract that fixes the price weeks later from published assessments, and keeps whatever survives after freight, financing, demurrage and quality adjustments have taken their share.