Aller au contenu
Corporate & Investment Banking Mon parcours English

Accueil / Capital Markets & Securities Services / Capital Markets / Module 10

10Fixed Income Trading

Execution & Trading · How banks make markets in the world's largest asset class: government and corporate bonds, the curve, repo and spreads, and the electronification remaking the desk

Faire le marché des obligations, la plus grande classe d’actifs au monde.

Lecture
2 h 00
Sections
72
Questions de rappel
35
Questions de QCM
20
Simulateurs
1

Le texte de ce module est encore en anglais : la version française arrive. Les outils interactifs, eux, sont déjà en français.

Two threads run through everything

The business thread is the craft of making markets. A dealer quotes a price it must be willing to trade on, takes the resulting risk onto its book, and manages that risk until it can pass it on. Everything in fixed income — pricing, hedging, repo financing, inventory, risk limits — exists to serve that one function. Read the module as the answer to a single question: how does a bank provide immediacy to a client, and get paid for it without being run over?

The technology thread is that the desk is now half software product. Rates trading electronified first, and credit is catching up fast. For the long tail of small, liquid tickets, a human no longer sets the price — a pricing engine does, thousands of times a second. The hard problems are data problems: many bonds do not trade on a given day, so their "price" must be modelled. Where the business thread asks how a dealer makes a market, the technology thread asks what it takes to make that market electronically, at scale, under tight capital limits.

How to read this module

The module runs in five parts. Part 1 builds the market and its products from first principles. Part 2 follows a trade from a client enquiry to a settled, financed position, stopping at every mechanic along the way. Part 3 maps the people and institutions. Part 4 opens up the systems and the data that connect them. Part 5 covers the risks, the failures, and the transformation agenda, then distils the whole module into six ideas. A one-page cheat sheet and a full glossary close the document.

Throughout, coloured boxes carry specific kinds of content. A Key insight box crystallises the single most important takeaway of a section. A Going deeper box adds nuance for readers who want the next layer. An IT & data lens box shows how a concept lives in systems and data. A Watch out box flags a trap or a moving target. A Worked example box runs the numbers. An In a nutshell box recaps a part, and a Check your understanding box lets you test yourself before moving on.

One running example threads through the module: a EUR 5 million corporate-bond request-for-quote from an asset manager, followed from the client's enquiry to a hedged, financed, reported position. A second, sharper example — a France-versus-Germany relative-value trade in government bonds — appears where the mechanics of rates trading are clearest. Both are illustrative; every figure attached to them is chosen to teach, not to report a real deal.

You will understand the fixed income market end to end

You will be able to explain why bonds trade over the counter rather than on an exchange, what a dealer actually does when it "makes a market", and how the two halves of the business — rates and credit — differ in liquidity, risk and technology. The market will stop being a wall of acronyms and become a machine you can reason about.

You will be able to follow a trade from enquiry to settlement

You will trace the seven steps every fixed income trade follows — enquiry, pricing, quoting, execution, hedging, clearing, settlement and financing — and understand why, unlike equities, the risk lives on the dealer's book in between. You will know where pricing, hedging and financing sit, and why they are the heart of the lifecycle.

You will understand where the price actually comes from

You will be able to build a two-way price yourself: start from a fair mid, add a spread for size and illiquidity, skew for the desk's axe and inventory, and factor in the hedge and funding cost. You will understand the credit spread as the single number a credit trader warehouses, and the yield curve as the market's read on the economy.

You will be able to trace data front to back

You will follow one trade through pricing, execution, risk, clearing, settlement and regulatory reporting, and you will see why an error at the front reaches every system behind it. You will know the difference between the FIX world of the front office and the SWIFT world of post-trade, and why bond data is uniquely hard.

You will recognise the risk and transformation landscape

You will be able to name the risks a dealer runs — market, spread, inventory, funding, counterparty and conduct — and the defences against each. You will understand why post-crisis capital rules made dealers safer but scarcer, and you will be able to judge which technology trends genuinely change the economics of the business and which merely change its appearance.

1The Market and the Big Picture

Before any trade, understand the terrain: what a bond is, why its market has no exchange, and why the dealer is indispensable.

Fixed income is the largest market on earth, and almost nobody outside it can say how it works. Equities get the headlines — a share price is a number anyone can quote — but the machinery that sets the price of money, funds every government and every large company, and underpins the valuation of every other asset is the bond market, and it runs on a completely different logic. This part builds that logic from zero. By the end you will understand what trades, why it trades the way it does, and why a bank standing in the middle is not a middleman skimming a fee but the thing that makes the market exist.

A bond is a loan you can sell, and that changes everything

A bond is a tradable loan. When a government or a company needs to borrow, it can go to a bank for a loan, or it can issue a bond: a security that promises to pay the holder a fixed stream of interest — the coupon — and to return the principal on a set date, the maturity. The difference between a loan and a bond is that a bond can be sold. The original lender does not have to wait years to get its money back; it can sell the bond to someone else at today's price. That single feature — transferability — is what creates a trading market, and everything in this module follows from it.

The price of a bond is just the present value of its promised cash flows. A bond that pays 3% a year and returns 100 at maturity is worth 100 only if 3% is exactly the return the market demands for that risk and maturity. If the market now demands 4%, the bond's fixed 3% coupon looks stingy, so its price falls below 100 until the return to a new buyer rises to 4%. This is the iron law of fixed income: when required yields rise, bond prices fall, and vice versa. Price and yield are two ways of stating the same thing, moving in opposite directions.

Yield is built from a risk-free base plus a premium for risk. The yield on a government bond from a solid sovereign is treated as the risk-free rate — the return you can get with near-certainty of repayment. Every other bond yields more, and the extra is the credit spread: the compensation investors demand for the chance that the issuer defaults, plus a premium for the fact that the bond may be hard to sell. A corporate bond's yield is therefore the risk-free rate plus its spread. Hold that decomposition in mind; it is the backbone of the entire credit business.

Coupon, maturity and the shape of a bond

A single issuer rarely has one bond. A large company or government issues many — different maturities, different coupons, sometimes different currencies and different levels of seniority. Each of these is a separate line with its own price. One equity, by contrast, is a single share. This asymmetry — one issuer, dozens of bond lines — is the root cause of the fragmentation that defines the whole market, and we return to it repeatedly.

The bond maths you actually need, built from zero

Before trading a bond you have to be able to measure it, and the measures sound intimidating but each is simple once built from scratch. Duration, DV01, convexity, clean and dirty price — these are the everyday vocabulary of the desk, and a reader who has never met them can understand every one from first principles. This section builds them, because the rest of the module leans on them constantly, and because nothing in trading makes sense until you can say what a price move actually does to a position.

Clean price, dirty price and the coupon that is always ticking

A bond pays its coupon on fixed dates, but it trades on every day in between — so on any given day the buyer owes the seller the slice of the next coupon that has accrued since the last payment. That slice is the accrued interest. The price you see quoted on a screen is the clean price, which strips the accrued interest out so that the number does not saw up and down as the coupon builds. The price actually paid at settlement is the dirty price: clean price plus accrued interest. If a bond has a 4% annual coupon and it is a quarter of the way to the next payment, roughly 1% of accrued interest is added to the clean price to get what changes hands. The distinction matters in every system in the module: traders quote and risk-manage on the clean price, but settlement moves the dirty price, and confusing the two is a classic booking error.

Duration — what a bond's price move really depends on

Duration is the single most important number about a bond: it tells you how much its price moves when yields move. The intuition is worth getting exactly right. A bond is a bundle of future cash flows — the coupons and the final principal — and duration is, in effect, the weighted-average time until you receive them, measured in years. A bond that pays everything back in two years has a short duration; a thirty-year bond that pays most of its value at the very end has a long one. The reason this "average time" matters is that the further away a cash flow is, the more its present value is affected by a change in the discount rate. So duration doubles as a sensitivity: a bond with a duration of seven years loses roughly 7% of its price if yields rise by one percentage point, and gains roughly 7% if they fall. Longer duration means a more violent reaction to the same yield move. That is why a rates trader thinks in duration before anything else.

Figure 1 Price falls as yield rises along a convex curve — duration is the slope, convexity is the curvature
DV01 — duration expressed in money

DV01 is duration turned into a currency amount, and it is how a desk actually manages risk. The name says it: the Dollar (or Euro) Value of one basis point — how much money a position gains or loses if yields move by a single basis point (one hundredth of one percent). It is the practical form of duration because a trader does not hedge "seven years of duration"; the trader hedges "forty thousand euros per basis point". The worked example below shows how it is computed and why it is the number on the screen.

Convexity — why duration is only an approximation

Duration assumes the price-yield relationship is a straight line, but it is actually a curve, and convexity is the name for that curvature. Look again at the diagram: the true relationship between a bond's price and its yield bends, while duration is the straight tangent line touching it at today's yield. For small yield moves the straight line is a fine approximation. For large moves it drifts away from the truth — and always in the bondholder's favour: prices rise a little more than duration predicts when yields fall, and fall a little less than predicted when yields rise. That helpful gap is convexity. It is a second-order effect, invisible on a quiet day and important on a violent one, which is why it matters most exactly when the market is moving hardest. For a credit trader the equivalent sensitivity to the credit spread is CS01 — the same idea applied to spread rather than yield — and Part 2 treats the two together.

How a bond is quoted — price, yield or spread

The same bond can be talked about in three languages, and knowing which one is being spoken avoids costly confusion. A government bond is usually quoted in price or yield, because it has a clean reference curve. A corporate bond is more often quoted in spread — its yield expressed as a number of basis points over the government curve or over swaps — because what a credit trader is really trading is the spread, not the outright yield. Two refinements are worth naming: the z-spread is the constant spread over the whole government curve that makes a bond's cash flows price correctly, and the asset swap spread is what you earn over the floating rate if you buy the bond and swap its fixed coupon to floating. Both are ways of isolating the pure credit compensation from the moving rate underneath, which is exactly the separation of rate and spread risk that runs through the whole module.

Credit ratings — the shorthand for default risk

A credit rating is a rating agency's opinion of how likely an issuer is to default, and it draws the single most important line in the credit market. The major agencies grade issuers on a scale from the safest (triple-A) down through the alphabet. Everything at BBB− and above is investment grade (IG): lower default risk, tighter spreads, a broad natural buyer base, and more liquidity. Everything below is high yield (HY), or "junk": higher default risk, wider spreads, more idiosyncratic and less liquid. The line between BBB− and BB+ is not just a notch on a scale — it is a cliff, because many investors are mandated to hold only investment grade, so a downgrade across that line can force a wave of selling. Ratings are an input, not gospel; a good credit trader forms their own view and treats the agencies' opinion as one data point among many.

Trading is the secondary market — where bonds live after they are born

There are two distinct moments in a bond's life, and this module is about the second one. In the primary market, a new bond is created and sold to investors for the first time; that is the debt capital markets (DCM) business, the bond equivalent of an IPO. In the secondary market, that bond is bought and sold over and over for the rest of its life. Fixed income trading is the secondary market. It is where a bond's price is discovered continuously, and where an investor who bought at issue can sell before maturity.

A liquid secondary market makes primary issuance cheaper, which is why the two are joined at the hip. Investors will pay more for a bond — accept a lower yield — if they are confident they can sell it later without a large discount. That confidence comes from knowing dealers will make a market in it. So the secondary desk's willingness to quote feeds directly back into how cheaply the issuer could borrow in the first place. Trading and origination are two ends of one business, and a bank that is strong in secondary market-making wins more primary mandates as a result.

Figure 2 The fixed income trading desk sits between issuers, the buy-side, venues, clearing and the official sector

Why there is no exchange, and why that is the whole story

Equities match on an exchange; bonds do not, and the reason is fragmentation. An exchange works when there is a continuous, two-sided flow of buyers and sellers for the same instrument, so their orders can be matched against each other in a central order book. That holds for a company's single share. It fails for its bonds. With dozens of lines per issuer and millions of individual bonds worldwide, most of them trade rarely — a given bond might not change hands for days or weeks. There is simply no continuous pool of resting buy and sell orders to match. Someone has to stand in the middle and quote a price on demand, absorbing the imbalance when there is a buyer but no seller, or a seller but no buyer.

That someone is the dealer, and quoting a two-way price is what "making a market" means. The dealer shows a bid — the price at which it will buy the bond — and an offer, the price at which it will sell. It must stand ready to trade either side. The gap between them, the bid-offer spread, is its raw compensation for providing the service and for holding the risk in between. This is a quote-driven market, not an order-driven one: the price comes from a dealer's quote, not from two clients meeting in an order book.

Two markets under one roof: rates and credit

Fixed income splits into two halves that share a name but little else. Rates desks trade government bonds, bond futures and interest-rate swaps. Credit desks trade corporate bonds and the derivatives that hedge them. Both are "fixed income", but the craft, the risk and the technology differ so much that they are run as separate businesses with separate traders, systems and cultures. Understanding the split is the first step to understanding the desk.

Rates — the deepest, most liquid market there is

The rates market trades sovereign debt: US Treasuries, German Bunds, French OATs, UK Gilts. These are the safest and most liquid instruments in the world, and their yields form the risk-free benchmark against which everything else is priced. Rates trading is enormous in volume, razor-thin in spread, and heavily electronic. The rates trader's mental world is the yield curve — yields plotted across maturities from a few months to thirty years — and its shape, which is the market's live read on growth and central-bank policy. Alongside the cash bonds sit deep, liquid futures and interest-rate swaps that let the desk hedge risk precisely and cheaply.

Credit — where you are paid to take default risk

The credit market trades corporate bonds: debt issued by companies and banks to fund themselves. Here the defining number is the credit spread — the extra yield over the risk-free rate that investors demand for taking the issuer's default and liquidity risk. Credit splits again into investment grade (higher-rated, tighter spreads, more liquid) and high yield (sub-investment-grade "junk", wider spreads, more idiosyncratic, far less liquid). Credit is less liquid than rates, carries higher margins, and is an inventory business: much of the value is simply in finding a bond a client wants when it rarely trades. Credit electronified later than rates, but is catching up fast.

Figure 3 The yield curve's shape — steep, flat or inverted — is the market's read on growth and policy

Three corners of the market worth naming before the product map

Beyond the plain government and corporate bond sit a few product families that round out the picture, and each answers a specific need. They belong in the map because a real desk trades them daily, and because each one isolates a different risk the plain bond bundles together.

The short end — money markets and Treasury bills

Not all debt is long-dated. At the very short end sit instruments that mature in days to a year or so: Treasury bills (short-dated government paper issued at a discount rather than paying a coupon), commercial paper (short-term corporate borrowing), and the money-market instruments banks use to manage day-to-day liquidity. This is the money market, and it matters for two reasons. It is where the central bank's policy rate transmits most directly into market rates, so it anchors the short end of the curve. And it is deeply tied to repo, because financing is itself a money-market activity — the repo rate lives here. A desk that trades the short end is trading the price of money over the next few weeks and months.

The inflation complex — linkers and real yields

Some government bonds pay a coupon and principal that rise with inflation; these are inflation-linked bonds, or "linkers" (such as index-linked Gilts, TIPS in the US, or OATi in France). They separate a risk the ordinary bond blends in: the ordinary bond's fixed coupon is exposed to inflation eroding its real value, while a linker protects against exactly that. Trading them means thinking in real yields (the return after inflation) rather than nominal yields, and the gap between a nominal bond's yield and its inflation-linked sibling — the breakeven inflation rate — is itself a closely watched, tradable measure of what the market expects inflation to be. Linkers are a smaller, less liquid corner than nominal govvies, but they are central to how the market prices inflation.

Rate derivatives — beyond the vanilla swap

The rates complex does not stop at cash bonds, futures and vanilla swaps. Beyond them lie interest-rate options — swaptions (options to enter a swap), caps and floors (options on a rate exceeding or falling below a level) — that let a desk express views on, and hedge, the volatility of rates rather than just their direction. These optional instruments carry the second-order risks (convexity and volatility) that the earlier bond-maths section introduced, and pricing them well is a quant-heavy discipline in its own right. This module treats them only as part of the landscape; the mechanics of options and structured rate products belong to the dedicated rates and derivatives module in the series. What matters here is knowing they exist and why a desk reaches for them: to trade and hedge the shape and volatility of rates, not merely their level.

The products, from the simplest and fastest to the most complex

Any fixed income desk trades a spectrum of instruments, and it helps to sort them by liquidity and speed. At one end sit deep, standardised, electronic instruments that trade in milliseconds at tiny spreads; at the other sit bespoke, illiquid positions that take skill and balance sheet to move. The table below lays out the main products across that spectrum. It is the map of the whole trading floor.

Product What it is Liquidity & speed Main risk Key technology
Government bonds (govvies) Sovereign debt — Treasuries, Bunds, OATs Deepest, near-instant, tightest spreads Rate / DV01 Streaming e-pricing, futures hedging
Bond futures Exchange-traded contracts on govvies Extremely liquid, exchange-matched Rate / basis Exchange connectivity, low latency
Interest-rate swaps OTC fixed-for-floating rate exchange Very liquid (vanilla), cleared Rate / curve Swap pricing, EMIR clearing
Investment-grade credit Higher-rated corporate bonds Moderately liquid, RFQ-driven Spread (CS01) + rate RFQ platforms, composite pricing
High-yield credit Sub-investment-grade corporate bonds Illiquid, often voice, high-touch Spread + idiosyncratic Inventory tools, axe distribution
Credit derivatives (CDS / indices) Synthetic credit exposure — iTraxx, CDX Liquid at index level Spread / basis Index pricing, cleared derivatives
Repo / securities financing Secured lending against collateral Huge volume, low margin Funding / counterparty Collateral optimisation, SFTR reporting

The reference case for this module is credit market-making — quoting and warehousing a corporate bond — because it exercises every muscle in the business. A government bond trade is mostly about speed and hedging; a corporate bond trade forces the dealer to model a price where none is quoted, take spread risk it cannot perfectly hedge, source a scarce instrument, and finance the position for as long as it holds. If you can follow a corporate bond from a client request to a settled, financed position — which is exactly what Part 2 does — you can follow anything else on the floor. The rates relative-value trade appears alongside it as the cleaner, more surgical counterpart.

Repo is the plumbing under everything

You cannot hold a bond for free, and repo is how the whole book is funded. A repurchase agreement, or repo, is a secured loan against a bond: the dealer sells a bond for cash today and agrees to buy it back shortly at a slightly higher price. Economically it is borrowing cash using the bond as collateral, and the price difference is the interest — the repo rate. The reverse — lending cash and taking a bond as collateral — is a reverse repo, and it is how a dealer borrows a specific bond it needs to deliver on a short sale. Every long position on the book is financed in repo; every short is covered through reverse repo. Financing is not back-office plumbing you can ignore. It is inseparable from trading, and Part 2 treats it as a business in its own right.

Why bonds are less liquid than equities — and why that makes the dealer indispensable

Liquidity is the ease of trading in size without moving the price, and in bonds it is structurally thin for four reasons. The first is fragmentation, already met: one company has a single share but can have dozens of bond lines, so whatever trading interest exists is spread thin across many instruments. The second is that bonds mature — they roll down toward their repayment date and eventually expire, so most trade actively only when freshly issued, and older lines go quiet ("off-the-run"). The third is that many holders buy to keep: insurers and pension funds hold bonds to maturity to match their liabilities, taking them out of circulation for years. The fourth is the absence of a central order book — without an exchange there is no continuous pool of resting buy and sell orders to match against. Put together, these mean that on any given day most bonds simply do not trade.

Because liquidity is thin, the dealer's willingness to quote is the liquidity — and that willingness is fragile. In calm markets a dealer will happily warehouse risk and quote tight, so the market feels liquid. In stress the exit disappears: everyone wants to sell the same way at once, the dealer cannot lay off what it takes on, and it either quotes far wider or steps back entirely. Liquidity is therefore state-dependent — abundant when it is not needed and scarce exactly when it is — which is a structural fragility of the whole market, not a flaw in any one dealer.

How the market is sized, and how league tables bend behaviour

Global bond markets dwarf equities, and fixed income (within the wider FICC franchise) is a core revenue engine for every large bank. The precise figures move constantly and depend on definitions, but the order of magnitude is not in doubt: the outstanding stock of bonds worldwide runs to well over one hundred trillion US dollars, far larger than the total value of listed equities. Governments and companies fund themselves overwhelmingly with debt, and every one of those bonds may need to trade. That scale is why the desk matters to the bank and why the technology spend on it is so large.

League tables rank dealers by volume, and what gets measured gets gamed. Independent data providers publish rankings of the most active dealers by traded volume and by market share on the major platforms. A high ranking wins credibility with clients and feeds back into primary mandates. The predictable distortion is that desks may chase volume — quoting aggressively on low-margin flow — to protect a ranking, even when the marginal trade barely covers its balance-sheet cost. A recurring theme of this module is that raw volume is a vanity metric; the return on the scarce resource, balance sheet, is what actually matters.

Why this business is cyclical, and why the cycle drives the technology

Fixed income trading breathes with the interest-rate cycle, and the last few years were a violent inhale and exhale. When central banks hold rates near zero and markets are calm, volatility is low, spreads are tight, and trading revenues are modest — there is little for the market to react to. When rates move sharply, volatility spikes, spreads gap around, and clients trade heavily to reposition — and dealer revenues surge. The years from 2021 to 2024 delivered exactly this: a long period of near-zero rates gave way to the fastest tightening cycle in decades as central banks fought inflation, and fixed income trading revenues jumped as volatility and client activity exploded. Then, as rates plateaued, conditions began to normalise.

The cycle matters for technology because it sets the budget and the priorities. Fat years fund investment; lean years force efficiency. But the deeper point is that each surge in volatility exposes exactly where the plumbing is weak — where pricing breaks down for illiquid names, where post-trade processing jams, where risk systems cannot keep up. The transformation agenda in Part 5 is, in large part, a list of the pressure points the recent cycle laid bare. A business that is calm most of the time but occasionally has to price and process enormous volumes under stress must build for the stress, not the calm.

Figure 4 Fixed income revenue tracks volatility — the business breathes with the rate cycle (illustrative)

The practical lesson of the cycle is that capacity must be sized for the peak, not the average. A desk that staffs, provisions and builds for a normal day will be overwhelmed on the day that matters — the day a central bank surprises the market, or a sovereign is downgraded, and volumes multiply within hours. That is why resilience, automation and scalable pricing are not luxuries funded in good years but the very things that determine whether the business captures the revenue of a volatile year or chokes on it. The cycle rewards the firms that built for stress while the market was calm, and punishes those that did not.

  1. Why do bonds trade over the counter through dealers rather than on a central exchange like equities?

  2. A bond pays a fixed 3% coupon. The market's required yield rises to 4%. What happens to the bond's price, and why?

  3. Decompose the yield on a corporate bond into its two parts. — The risk-free rate (from the government curve) plus the credit spread, which compensates for default risk and liquidity risk.

  4. What is the difference between the primary and secondary bond markets, and how are they linked?

  5. How do rates and credit differ as trading businesses?

  6. Why is liquidity in bonds described as state-dependent, and what limits how much a dealer can provide?

  7. Why is raw traded volume a misleading measure of a desk's success?

  8. In one sentence each, what are duration and DV01?

1

  1. Why do bonds trade over the counter through dealers rather than on an exchange?

  2. A bond pays a fixed 3 % coupon and the required yield rises to 4 %. What happens?

  3. What are the two parts of a corporate bond's yield?

  4. Why does a liquid secondary market lower issuers' borrowing costs?

2From Enquiry to a Settled, Financed Position

This is the heart of the module: the path every fixed income trade follows, and the mechanics that live at each step.

A fixed income trade looks, from the outside, like a single click. An asset manager wants a bond, asks a few dealers for a price, takes the best one, and the trade is done. Underneath that click is a chain of decisions that determines whether the dealer makes money or loses it: how to price the risk, whether to skew the quote, how to hedge what is left, how to finance the position, and how to move it safely through clearing and settlement. This part walks the whole chain, one step at a time, and stops at every mechanic that deserves a closer look. It is deliberately the longest part of the module, because this is where the craft lives.

The seven steps every trade follows

Every fixed income trade travels the same seven steps, from a client enquiry to a settled and financed position. The names change between desks, but the sequence does not. A client asks for a price; the dealer prices the risk; it shows a quote and the client accepts or passes; the trade is executed and booked; the desk hedges and manages the resulting risk; the trade is confirmed and, where applicable, cleared; and finally it settles and the position is financed. The table below is the map for the rest of Part 2.

# Stage Typical duration What is produced Who leads
1 Enquiry (RFQ or voice) Seconds A request for a price on a specific bond Client, via sales / platform
2 Price the risk Milliseconds to minutes A fair mid and a bid / offer Trader or pricing engine
3 Quote & negotiate Seconds A firm two-way or one-way price Trader / auto-quoter
4 Execute & book Instant An agreed trade captured in the risk system Trader; system
5 Hedge & manage Minutes to days A hedged position within limits Trader
6 Confirm & clear Same day (T+0) An agreed, cleared trade Operations; CCP
7 Settle & finance T+1 / T+2, then ongoing Bond and cash exchanged; position funded Operations; repo desk

The effort and the risk are not spread evenly across the seven steps — they spike in the middle. The first step is trivial; the last steps are largely automated plumbing. The danger concentrates where the dealer commits to a price and takes the risk onto its book: steps two through five. That is where a mispriced quote, a missed hedge, or a stale position turns into a loss. The intensity curve below shows where the pressure sits, and it is the reason the rest of this part dwells on pricing, hedging and financing far more than on enquiry or settlement.

Figure 5 The fixed income trade lifecycle — effort and risk peak where the dealer prices and takes the risk

The running example: a EUR 5 million corporate-bond RFQ

We will follow one trade through the whole part, so the mechanics attach to something concrete. The figures are illustrative — chosen to teach the mechanics, not to report a real transaction — but the flow is exactly how a modern credit trade works.

Before any trade — the pre-trade gauntlet

A client cannot simply appear and trade; a gauntlet of internal checks stands before the first quote, and it runs largely in the background. The trade lifecycle described above assumes a client the bank has already onboarded and a product the bank is already approved to trade. Getting to that point is its own controlled process, and it is worth understanding because it is where a great deal of the technology and the risk actually sits.

The first gate is client onboarding and know-your-customer (KYC). Before a client trades, the bank must verify who it is, assess it for financial-crime and sanctions risk, and set it up in every system that will touch its trades. This is a heavy, data-intensive process, and it is a genuine gate: a client that is not fully onboarded cannot be traded with, no matter how attractive the flow. Onboarding also establishes the client's credit standing, which feeds the next gate.

The second gate is the credit limit, and it is checked before every trade, automatically. Because the dealer takes principal risk and may be owed money between trade and settlement, it extends the client a credit line — a limit on how much exposure it will carry to that counterparty. The pricing engine checks the proposed trade against the available credit line in real time; a trade that would breach the limit is blocked or escalated. This pre-trade limit check is invisible when it passes and decisive when it does not, and it is one reason clean, current counterparty data matters so much.

The third gate is new-product approval, the internal committee that must bless a product before the desk can trade it. A bank does not let a desk trade a novel or complex instrument on a trader's say-so. A new-product-approval process — a committee spanning trading, risk, legal, compliance, operations and technology — assesses whether the bank can price it, risk-manage it, book it, settle it and report it before it is allowed. This is the trading equivalent of the internal governance that gates any major commitment: the point is to ensure the whole machine, front to back, can actually handle the product before the first trade, not to discover the gaps afterwards.

Step 1 — The client asks, and the channel already shapes the trade

A trade begins when a client asks the desk for a price, and how it asks matters. The enquiry arrives by voice (a phone call or a chat message) or, increasingly, as an electronic request-for-quote on a platform. Voice still dominates for large, complex or illiquid trades, where the client wants discretion and a human conversation about size and level. Electronic RFQ dominates for smaller, more liquid tickets, where speed and competition matter more than discretion. In our example the asset manager uses an electronic RFQ to three dealers — a competitive auction that will be decided in milliseconds.

The number of dealers asked, and whether they can see each other, changes the dynamics. An RFQ to three dealers is a sealed auction: each dealer quotes without seeing the others, and the best price wins. Ask too few dealers and the client may not get the keenest price; ask too many and dealers quote defensively, knowing their chance of winning is low and their quote may leak information about the client's intention. This tension — competition versus information leakage — runs through every marketing and execution decision in fixed income, and we meet it again in the discussion of pre-sounding and anonymity below.

Step 2 — Pricing the risk: how a two-way price is actually built

Pricing is the intellectual core of the business, and it follows a repeatable recipe. The dealer (or its pricing engine) builds a price in four moves: start from a fair mid, add a spread, skew for inventory, and factor in the hedge and funding cost. Each move is a distinct judgement, and getting any of them wrong loses money.

Move one — find the fair mid

The fair mid is the dealer's best estimate of the bond's true price right now. For a liquid government bond, the mid is almost given — the market quotes it continuously. For a corporate bond that has not traded in days, the mid must be modelled: the dealer starts from the government curve, adds the spread implied by comparable bonds from the same issuer or sector, and cross-checks against CDS, index and ETF levels. This is the fixed income equivalent of valuation, and it is genuinely hard — a mistake here poisons everything downstream.

Figure 6 Building a fair mid — triangulating the curve, comparables, synthetic spreads and the desk's own axe
Move two — add a spread for size, illiquidity and risk

Around the fair mid the dealer places a bid and an offer. The width of that spread reflects how much risk the trade puts on the book: wider for large size, for illiquid names, and for volatile conditions; tighter for small, liquid, easy-to-hedge trades and for valued clients. The spread is the reward for immediacy — for standing ready to trade now — and it must at least cover the expected cost of hedging and holding the position until it can be offloaded.

Move three — skew for the axe

This is where the dealer's own inventory bends the price. If the desk is long a bond it wants to sell, it quotes a keener offer to attract buyers; if it is short and wants to buy, it quotes a keener bid. The desk's axe — its desire to do a particular trade — shows up directly in the price. In our example the dealer is short and axed to buy, so it will show a sharp offer to win the client's buy and flatten its short. The trade that reduces your risk is the trade you price to win; the trade that adds risk is the trade you price to lose, or to be well paid for.

Move four — factor in the hedge and funding cost

Finally the dealer folds in what it will cost to hedge the new risk (with a future or a swap) and to finance the position in repo until it is sold. A bond that is expensive to fund, or awkward to hedge, must be priced wider to compensate. Only once all four moves are made does a firm, tradable price exist.

Step 3 — Quoting and negotiating: the moment of commitment

When the dealer shows a price, it is committing to trade on it, and on a platform the best price simply wins. In a voice trade the dealer shows a two-way or one-way price and the client accepts, passes, or negotiates on level or size. On an electronic RFQ there is no negotiation: each dealer's quote stands, the platform ranks them, and the client clicks the best. In our example the dealer's sharp offer — sharp because filling the client covers its short — is the best of the three, and the client clicks it. The EUR 5 million trade is agreed.

The dealer's win-rate is a managed number, not an accident. Win every RFQ and you are pricing too tight, giving risk away; win almost none and you are wasting the desk's quoting capacity and signalling nothing to clients. Desks track hit-rates by client, size and spread and tune their pricing to win the flow they actually want — flow that fits the book — while losing the flow that would only add unwanted risk. Quoting is therefore not about being cheapest; it is about being cheapest on the trades you want and dearest on the trades you do not.

Step 4 — Execution and booking: the risk lands on the book

The instant the trade is agreed, it is captured in the risk system and the dealer's position and P&L update. On an electronic platform this is automatic: the click books the trade, updates the position, and recalculates the desk's risk in real time. In our example the dealer's short position shrinks by EUR 5 million as the buy-back lands, and its risk measures move accordingly. This is the moment the dealer becomes a principal in the trade — it now owns a changed position, and everything that follows is about managing it.

Clean booking is not clerical detail; it is the foundation of everything downstream. The booked trade record — instrument, price, size, counterparty, timestamps — becomes the single source of truth for risk, P&L, confirmation, clearing, settlement and regulatory reporting. An error at booking propagates to every one of those systems. Part 4 returns to this as the data-lineage principle; for now, note that the trade capture in step four is the atom from which the whole back end is built.

Step 5 — Hedging and managing the position: isolating the risk you want

A booked position almost always carries risk the desk does not want, and hedging strips it away. A corporate bond carries two distinct risks: the risk that interest rates move (measured by DV01, the profit or loss for a one-basis-point move in yields) and the risk that the credit spread moves (measured by CS01, the profit or loss for a one-basis-point move in the spread). A credit desk usually wants the spread risk — that is its view and its franchise — but not the underlying rate risk. So it hedges the rate risk away with a liquid bond future or an interest-rate swap, leaving a cleaner spread position. What remains is the risk the desk chose to hold.

Figure 7 A corporate bond carries two risks — rate (DV01) and spread (CS01) — measured and hedged separately

Managing the position after the hedge is a continuous discipline, not a one-off. The desk watches its net risk against limits, looks for the other side (another client, the interdealer market, an ETF or index) to recycle the position and turn inventory over, and keeps an eye on concentration and age. Stale inventory — a position no one wants, sitting for weeks — ties up balance sheet and quietly signals that the original price was wrong. Good desks recycle fast; bad ones accumulate. The whole art of the flow model is to warehouse client flow, hedge the net, and move it on before it becomes a problem.

Deep dive — the hedging toolkit

Hedging is only possible because a handful of deep, liquid instruments let the desk offset specific risks cheaply and instantly, and knowing what each one does is part of understanding the trade. A hedge is not a vague act of caution; it is a precisely sized offsetting position that removes one named risk while leaving the rest of the book untouched. The five instruments below are the everyday tools.

Bond futures — the everyday rate hedge

Exchange-traded futures on government bonds (the Bund, OAT and Treasury futures) are extremely liquid and cheap to trade, which makes them the default tool for hedging rate risk. Sell futures against a long bond position and a rise in yields costs the bond roughly what it makes on the short future. Because futures are exchange-matched and margined, they are fast, transparent and easy to size by DV01 — the reason they are the first thing a trader reaches for.

Interest-rate swaps — hedging and positioning the curve

An interest-rate swap is an over-the-counter contract to exchange fixed for floating interest payments. It is the core hedging and positioning tool of the rates desk: a swap can hedge the rate risk of a bond, express a view on a specific part of the curve, or convert a bond's fixed coupon into a floating one (an "asset swap") to isolate its spread. Standardised swaps are centrally cleared under the EMIR obligation, which makes them safer to use in size.

CDS, indices and ETFs — hedging credit

Rate risk is easy to hedge; spread risk is harder, and three instruments do most of the work. A credit default swap (CDS) is a contract that pays out if a specific issuer defaults, letting a desk hedge the default risk of a single name synthetically. CDS indices — iTraxx in Europe, CDX in the US — bundle many names into one liquid instrument, so a desk can hedge broad credit risk in a single trade. And credit ETFs, increasingly, let a desk trade and hedge a whole basket of credit in one liquid, exchange-traded line. None of these perfectly offsets the spread risk of one specific bond, which is exactly why credit remains an inventory business — the residual, un-hedgeable spread risk is what the credit trader is paid to warehouse.

Figure 8 A credit default swap — the buyer pays a periodic premium; the seller pays out if the issuer defaults
The basis — where cash, derivatives and financing meet

The gap between related instruments is itself a tradable risk, and managing it is where hedging becomes a craft. A bond and its future, or a bond and its swap, should move together — but not perfectly, and the difference is the basis. A hedge built from futures or swaps leaves a small residual basis risk, and desks both manage that risk and, sometimes, trade it deliberately. The basis is also where financing intrudes: the cost of repo-funding the cash bond feeds directly into the cash-versus-future basis, which is why hedging, financing and relative value are not three separate topics but three views of one book.

Instrument Hedges Why it is used
Bond future Rate risk (DV01) Deep, liquid, exchange-matched, easy to size
Interest-rate swap Rate / curve risk Precise curve positioning; asset-swap to isolate spread
Single-name CDS Default risk of one issuer Synthetic, no need to hold or borrow the bond
CDS index (iTraxx/CDX) Broad credit / spread risk One liquid instrument for many names
Credit ETF Basket credit risk Exchange-traded, instant, used to price/hedge portfolios

Deep dive — market-making mechanics and inventory management

Making markets well is a craft that goes beyond any single trade, and it centres on managing the book as a whole. A dealer is not pricing trades in isolation; it is running an inventory, and every quote is shaped by what it already holds. The four disciplines below are what separate a desk that earns its spread from one that slowly bleeds it away.

Figure 9 The axe bends the quote — a long desk keens its offer to sell, a short desk keens its bid to buy

The first discipline is to keep net risk small. The modern flow model does not take large directional bets; it warehouses client flow and hedges the net, earning the bid-offer spread across huge volumes while keeping the overall position modest. The second is to recycle positions actively — to find the other side quickly rather than sitting on risk. The third is to watch concentration and age, avoiding too much of one name and refusing to let positions grow stale. The fourth is to publish axes: to broadcast to clients what the desk wants to do, so that matching flow comes to it. Publishing an axe turns inventory management into client service — the client gets a sharp price, the desk reduces its risk, and both are better off, exactly as in our running example.

Deep dive — how the desk makes money

Fixed income trading has several distinct revenue engines, and a strong desk runs all of them. They are worth separating because they carry very different risk profiles and depend on very different capabilities.

Revenue engine How it earns Risk profile
Bid-offer spread Buy at the bid, sell at the offer, capture the difference Low if hedged and recycled; the core flow business
Carry Earn a bond's yield while paying a lower repo funding cost Moderate; depends on funding staying cheap
Financing / repo Run a matched book — lend and borrow cash against collateral for a spread Low margin, high volume, balance-sheet intensive
Positioning / views Take directional or relative-value positions that profit if a view is right High risk, high reward; constrained post-crisis
Hedged flow franchise Warehouse and hedge huge client volumes, net risk kept small Low net risk; the modern core model
New-issue / primary link Make secondary markets in newly issued bonds, feeding DCM franchise Low; franchise value more than P&L

The industry has shifted decisively from the high-risk engines to the low-risk ones. Directional risk-taking has shrunk, partly by regulation and partly by economics, and the hedged flow franchise has become the core. This matters for technology: a flow business that lives on volume and thin spreads can only be profitable if it is efficient — if pricing is automated, hedging is cheap and instant, financing is optimised, and balance sheet is used with discipline. The revenue model and the technology agenda are the same story told twice.

Deep dive — repo and financing as a trading business

Financing is not the plumbing beneath trading; it is a trading business in its own right, and it is the constraint on everything else. The financing (or repo) desk has four jobs. It funds the firm's entire bond inventory as cheaply as possible. It runs a matched book — borrowing and lending cash and collateral, earning the spread between the two sides. It sources specials — bonds in high demand that finance cheaply, which is a real profit centre. And it optimises collateral, delivering the cheapest acceptable collateral and managing haircuts to minimise the cost of the whole book.

General collateral versus specials

Most financing is done at the general collateral (GC) rate — the generic market rate for lending cash against any acceptable bond. But some bonds are in high demand to borrow — perhaps everyone wants to short them, or they are needed to deliver into futures — and these trade "special": they finance more cheaply than GC because the lender of the bond is willing to accept a lower rate on the cash to get hold of the security. Spotting and lending specials is where the financing desk adds value beyond simple funding.

Carry — the gap that quietly drives the P&L

Carry is the gap between what a bond yields and what it costs to fund. If a bond yields 4% and the desk funds it in repo at 3%, it earns 1% a year in carry simply for holding the position — money that accrues daily whether or not the price moves. Positive carry is a tailwind; negative carry (funding cost above the yield) is a drag that forces the desk to be right on price quickly. Carry is a real and often underappreciated driver of trading profit and loss, and it is why the repo rate — the cost of holding — sits at the centre of every position decision.

Figure 10 Repo mechanics — the dealer delivers a bond, receives cash today, and reverses it later with interest

Deep dive — measuring the book: DV01, CS01, VaR and limits

You cannot manage what you cannot measure, and a fixed income desk lives by a small set of risk numbers. These are the instruments on the dashboard, watched continuously and bounded by hard limits.

Measure What it captures Which desk cares most
DV01 / duration P&L for a 1bp move in yields — the primary rate risk Rates (and rate-hedged credit)
CS01 P&L for a 1bp move in the credit spread — the primary credit risk Credit
Convexity / gamma How DV01 itself changes as yields move — second-order risk Rates, options desks
VaR & expected shortfall Statistical estimate of potential loss over a horizon at a confidence level Firm-wide risk limit
Scenario / stress tests The book's P&L under specific shocks (rate spike, spread blowout) Risk management

The daily discipline is to separate the two first-order risks and manage each on purpose. A corporate bond carries both DV01 (rate) and CS01 (spread) risk. A rates desk hedges away everything except the specific rate view it wants; a credit desk hedges away the rate risk to be left with pure spread risk. Above these sits Value-at-Risk — a statistical estimate of how much the whole book could lose over a given horizon at a given confidence level — which serves as the firm-wide limit metric, backed by scenario and stress tests that capture the extreme moves VaR alone can miss. Every desk trades inside DV01, CS01, VaR and concentration limits, monitored independently by a market-risk function that does not report to the traders. That independence is the control that keeps a good desk honest.

Deep dive — the electronic market: how flow actually reaches the desk

The channel a trade travels is not neutral plumbing; each protocol shapes who competes, what leaks, and how the risk moves. Fixed income electronified from the top down — the most liquid, standardised instruments first — and the protocols that carry the flow are worth knowing individually, because Part 4's technology and Part 5's transformation both build directly on them.

Dealer-to-client and dealer-to-dealer

The market has two layers. In the dealer-to-client (D2C) layer, clients request prices from dealers on platforms such as Tradeweb, MarketAxess and Bloomberg — this is the main electronic channel for the buy-side. In the dealer-to-dealer (D2D) layer, dealers trade with each other to offload risk, often anonymously, on interdealer platforms and central order books. A trade a dealer wins from a client in the D2C layer is frequently laid off, in whole or in part, in the D2D layer — the two layers are how risk enters the dealer's book and how it leaves.

RFQ and streaming — the two workhorse protocols

The request-for-quote (RFQ) protocol is the workhorse of bonds: a client asks one or several dealers for a price on a specific bond and trades the best. It suits instruments that do not have a continuous price, which is most of them. Streaming (or click-to-trade) is the other model: dealers stream live, continuously updated prices that a client simply clicks to deal, which works only where the instrument is liquid enough to be priced continuously — government bonds, futures and vanilla swaps. Roughly speaking, the more standardised and liquid the instrument, the more it moves from RFQ toward streaming.

Figure 11 Electronification by product — rates led, credit is catching up (illustrative orders of magnitude)

Rates and credit sit at different points on the electronic journey, and the gap is closing. Government bonds, futures and vanilla swaps are heavily electronic — standardised, liquid and tight. Corporate bonds were long voice-only, but RFQ, portfolio trading and all-to-all are electronifying credit fast. The percentages move constantly and depend on definitions, so treat them as orders of magnitude; the direction, not the decimal, is the point.

Portfolio trading — moving a basket in one risk transfer

Portfolio trading lets a client buy or sell a whole basket of bonds — often dozens or hundreds of line items — in a single transaction against one dealer price. Its power is that the liquid bonds in the basket "carry" the illiquid ones: the client moves a large, mixed portfolio in one risk transfer that would be impossible to do bond-by-bond by voice. It is enabled by data and by credit ETFs, which let the dealer price and hedge the whole basket instantly. The result is fast, certain execution of large, diversified credit portfolios — one of the two innovations that let credit trade at scale despite thin dealer balance sheets.

All-to-all — when the buy-side becomes a price-maker

All-to-all trading breaks the old rule that only dealers quote. On an all-to-all protocol, asset managers, hedge funds and others can respond to any request, not just banks — so the buy-side becomes a price-maker, not only a price-taker. This expands the pool of liquidity and eases the market's reliance on dealer balance sheets, and it is often anonymous, which reduces information leakage. MarketAxess pioneered the model with its Open Trading platform, and it has become a structural shift in how corporate bonds change hands. Together, portfolio trading and all-to-all are the reason credit liquidity has held up even as dealer inventory shrank.

Deep dive — best execution and transaction cost analysis

When a bank executes a client's order it owes a duty of best execution, and in bonds that duty is genuinely hard to discharge. Under the European rules (MiFID II), a firm must take all sufficient steps to get the best possible result for the client, weighing price, cost, speed, likelihood of execution, size and settlement. In equities that is straightforward — there is a continuous reference price to measure against. In bonds there often is not: many instruments trade rarely, there is no single benchmark price, and for a large or illiquid trade, certainty of execution and low market impact can matter more than the headline level. The firm must nonetheless have an execution policy, monitor outcomes, and evidence best execution to clients and regulators.

Transaction cost analysis (TCA) is how execution quality is measured, and it is fundamentally a data problem. TCA grades each trade against benchmarks — composite prices, the cover (the second-best quote in an RFQ), the arrival price — to quantify how good the execution was. The buy-side uses TCA tTransaction cost analysis (TCA) is how execution quality is measured, and it is fundamentally a data problem. TCA grades each trade against benchmarks — composite prices, the cover (the second-best quote in an RFQ), the arrival price — to quantify how good the execution was. The buy-side uses TCA to grade its dealers; the sell-side uses it to prove and improve its pricing and hit-rates. The obstacle is data: bond TCA needs composite pricing and trade data pieced together across many venues and dealers, which is a serious data-engineering task. As electronic RFQ and post-trade transparency data improve — and as the EU builds a consolidated tape for bonds — TCA is steadily becoming more feasible, turning a regulatory duty into a competitive edge.

Deep dive — settlement, clearing and the close of the loop

After the trade is agreed, it must be confirmed, cleared, settled and financed — and this is where risk is really contained. The steps look like plumbing, but each one removes a specific danger.

Confirmation and affirmation come first: both sides agree the trade's economics, ideally electronically and the same day, so there is no dispute later about what was done. Clearing comes next where it applies: a central counterparty (CCP) steps in between buyer and seller, becoming the buyer to every seller and the seller to every buyer, netting exposures and guaranteeing the trade so that neither side is exposed to the other's default. Settlement is the moment bond and cash actually change hands, at a central securities depository (CSD), typically one or two business days after the trade (T+1 or T+2), on a delivery-versus-payment basis so that neither leg moves without the other. Finally, the resulting position is financed in repo until it is closed — which closes the loop back to the financing deep dive above.

What clears where

Not everything clears the same way, and the pattern matters. European government bonds and repo are heavily CCP-cleared through venues such as LCH and Eurex, which net enormous volumes and cut counterparty risk. Standardised interest-rate swaps must be centrally cleared under the EMIR clearing obligation, a post-2008 reform. Cash corporate bonds, by contrast, typically settle bilaterally through CSDs rather than through a CCP, because they are less standardised. Settlement discipline rules (CSDR) penalise failed settlements, pushing the whole market toward automation, electronic matching and shorter cycles.

Deep dive — the post-trade life of a position

A trade is not over when it settles; the position then lives on the book and has to be managed to its exit. In our example the desk bought the bond to cover a short, so the "position" is really the flattening of an earlier one — but in the general case the dealer now holds inventory it must eventually sell. Three things then matter. The first is turnover: how quickly the desk can recycle the position to another client, the interdealer market, or a hedge via an index or ETF. The second is carry: while it holds the bond, the desk earns the yield and pays the repo cost, and that gap accrues daily. The third is ageing and concentration: the longer a position sits and the larger any single name grows, the more balance sheet it consumes and the more it signals a pricing error. The discipline of the book is to keep inventory young, diversified and moving.

Deep dive — the interdealer market, the escape valve

When a dealer cannot find a client to take the other side, it turns to other dealers. The dealer-to-dealer (D2D) market — often anonymous, run through interdealer brokers and central order books — is where dealers offload risk to each other. It is the pressure-release valve of the whole system: a desk that has warehoused a position from a client but cannot recycle it into client flow can lay it off, in whole or in part, to another dealer who wants it. Anonymity matters here, because a dealer showing that it urgently needs to sell a large position would move the price against itself. The interdealer market is less visible than the client-facing business, but it is essential to how risk circulates and eventually clears.

Deep dive — the second example: a rates relative-value trade

The credit RFQ shows market-making; a rates relative-value trade shows the other half of the craft — isolating one risk and hedging everything else. Where the credit desk warehouses spread risk it cannot perfectly hedge, the rates desk uses deep, liquid futures and swaps to strip a position down to exactly the view it wants and nothing more. This surgical precision is why rates was the first fixed income market to electronify, and it is worth following one trade to see it.

The mechanics of the trade are a compact tour of everything in this part. First, size by DV01: match the DV01 of the OAT long and the Bund short so a parallel move in yields nets to zero. Second, finance both legs: the long OAT is funded in repo, while the short Bund is borrowed via reverse repo and delivered — and the carry on both legs is watched closely, because financing can quietly make or break a relative-value trade held for weeks. Third, isolate the risk: with the level and curve risk hedged out, what remains is pure OAT–Bund spread risk, exactly the view. Fourth, manage and monitor: track the spread daily against limits, trim or add as it moves, and watch for political, fiscal or supply catalysts that could move sovereign spreads sharply. If the spread tightens as expected, the trade profits with minimal outright rate risk — textbook relative value.

  1. Name the seven steps of the fixed income trade lifecycle and say where the risk concentrates.

  2. Walk through the four moves that build a two-way price.

  3. What does it mean that the desk is "axed to buy", and how does it change the quote?

  4. A credit desk buys a corporate bond. What two risks does it now hold, and how does it typically manage them?

  5. What is carry, and how can repo turn it from a tailwind into a headwind?

  6. How do portfolio trading and all-to-all help credit trade at scale despite thin dealer balance sheets?

  7. What does a CCP do, and why do corporate bonds usually not use one?

  8. In the OAT–Bund relative-value trade, why are the two legs sized to equal DV01?

2

  1. Where does risk concentrate in the fixed income trade lifecycle?

  2. Which move is NOT part of building a two-way price?

  3. The desk is “axed to buy”. What does it do?

  4. A credit desk buys a corporate bond. What risks does it hold?

3The Actors and How They Interact

A market is a set of relationships. This part maps who is involved, why the dealer sits in the middle of every one, and how the desk is organised inside the bank.

Part 2 followed a trade; this part follows the people. Fixed income trading connects a small number of distinct camps — those who issue debt, those who make markets in it, those who invest in it, and the infrastructure that clears and settles it — and the dealer touches every one. Understanding the actors is not a matter of memorising an org chart. It is understanding why each party trades, what each wants from the dealer, and where the frictions and the conflicts sit. That understanding is what lets you reason about the market rather than just name its parts.

Four camps, and the dealer in the middle of all of them

The market divides into four camps, and the dealer is one party to almost every interaction. The issuers create the bonds. The buy-side invests in them. The infrastructure clears and settles them. And the dealer — the sell-side bank — stands between all of them, making the market, warehousing the risk, and moving the trade through the machinery. This is the same hub-and-spoke shape as the actor map in Part 1, and it is worth stating plainly why the dealer sits at the centre: because in a market with no exchange, the dealer's willingness to quote and to hold risk is the only thing connecting a seller who wants to sell now with a buyer who will not appear until later.

The issuers are governments — issuing through debt agencies — and companies raising funding. They are the source of the bonds that trade, and although they operate mostly in the primary market, their choices about what and when to issue ripple straight into secondary supply and pricing. The buy-side is the investing world: asset managers, hedge funds, insurers and pension funds who buy, sell and, increasingly, provide liquidity themselves. The infrastructure is the venues, central counterparties and central securities depositories that connect, clear and settle the flow. And around all of them sit the data and analytics vendors who supply prices and indices, the interdealer brokers who match dealer-to-dealer flow, and the regulators and central banks who set the rules and move the rates market.

Interaction From To Stage it belongs to
Request for quote Buy-side client Dealer Enquiry
Two-way price Dealer Buy-side client Quote
Execution / fill Venue Dealer & client Execution
Risk offload Dealer Interdealer market / another dealer Hedge & manage
Hedge trade Dealer Futures exchange / swap CCP Hedge & manage
Clearing instruction Dealer CCP Clear
Settlement (DVP) Dealer & client CSD Settle
Financing (repo) Dealer Repo counterparty / CCP Finance
Prices & curves Data vendors Dealer Pricing (continuous)
Transaction report Dealer Regulator Post-trade (continuous)

The buy-side is not one thing — and the differences drive the desk

The single most useful distinction on the buy-side is why each investor trades, because it shapes how the dealer prices and allocates to them. A one-size-fits-all view of "clients" misses the entire point. A pension fund and a hedge fund want opposite things from the same bond, behave differently in stress, and are worth different amounts to the desk. The four archetypes below are the ones every trader carries in their head.

Real money — asset managers, the biggest natural clients

Real-money accounts are long-only funds, index trackers and pension mandates. They buy bonds for yield and diversification and tend to hold them, trading to adjust allocations rather than to speculate. They are the largest natural client base, they move in size, and they are relatively price-insensitive on any single trade because their horizon is long. A dealer prizes real-money flow: it is stable, it is large, and it is less likely to trade against the dealer's position immediately after.

Hedge funds — the active traders who also supply liquidity

Hedge funds trade for return: relative value, macro, credit long/short. They are active, sophisticated, and price-sensitive — they will work hard for a better level and will trade in and out quickly. Their fast, two-way flow adds liquidity to the market, and increasingly they supply liquidity directly through all-to-all protocols rather than only taking it from dealers. The flip side is flipping risk: a hedge fund may sell straight back what it just bought, so a dealer reads their flow carefully.

Insurers and pension funds — the patient long-end buyers

Insurers and pension funds buy long-dated bonds to match long-dated liabilities — the pensions they will pay and the claims they will meet decades out. They are large, patient, and famously price-insensitive at the long end of the curve, because their need is to match duration, not to time the market. They take bonds out of circulation for years, which is one reason older bonds go quiet. For a dealer, this is deep, reliable demand — but concentrated at the long maturities.

Retail — small individually, large in aggregate, politically sensitive

Retail investors buy bonds in small size, usually indirectly through funds and ETFs rather than directly. Individually they are a rounding error; in aggregate they are a meaningful and growing source of demand, and they carry a political dimension — how ordinary savers are treated in markets is a matter of public and regulatory attention. Retail flow reaches the desk mostly through the funds and platforms that pool it, not through direct RFQs.

Investor type Holding horizon Price sensitivity Why the dealer wants or watches them
Real money (asset managers) Long — buy and hold Low on a single trade Large, stable flow; the core client base
Hedge funds Short — trade actively High — works every basis point Two-way liquidity, but flipping risk to watch
Insurers / pension funds Very long — liability-matched Very low at the long end Deep, reliable long-end demand
Retail (via funds/ETFs) Mixed, mostly indirect Low individually Breadth and growth; politically sensitive

The official sector — the biggest force in the room

Central banks and debt agencies are not participants like the others; they set the terms everyone else trades within. A central bank sets the policy rate that anchors the short end of the curve, and through quantitative easing and tightening it buys and sells bonds in enormous size, moving the whole curve. In Europe the ECB is the dominant force in rates, and national debt agencies — such as Agence France Trésor for France — issue sovereign debt through a network of primary dealers that banks compete to belong to. When the official sector moves, the market moves with it; no private participant comes close to that weight. A rates trader who ignores central-bank policy is not trading rates.

Information barriers — one institution that must not know itself

A bank in fixed income knows things in one part of the house that another part must not use, and keeping them apart is both a legal duty and a technology problem. The classic case is the wall between primary and secondary, and between the bank's own risk-taking and its client business. The origination side may learn, in confidence, that a client is about to issue a large bond — material, non-public information. The trading desk must not trade on that knowledge, and the sales force must not leak it. The controlled functions are kept separate: those with confidential client information on one side, those trading in the market on the other, with formal, logged exceptions when someone must cross the wall.

Wall-crossing is the formal, controlled exception, not a loophole. When a person genuinely needs to be brought over the wall — to work on a confidential deal, say — it is done through a documented process: they are recorded as wall-crossed, restricted from trading the relevant names, and monitored. The point is that the barrier is auditable. A regulator or a surveillance system must be able to show who knew what, when, and whether they acted on it.

Inside the bank — the teams you will actually work with

A modern fixed income desk is a blend of traders, a financing function, an independent risk and compliance layer, and a fast-growing technology axis. It is now, only half-jokingly, half trading floor and half software product. The six roles below are the ones a newcomer meets, and mapping each to a stage and a system is the quickest way to see how the business hangs together.

Rates trader

Makes markets in government bonds, futures and swaps; manages DV01 and curve risk. High volume, tight spreads, deeply electronic. The rates trader's day is precision hedging and relative value — isolating a piece of the curve and stripping out everything else.

Credit trader

Makes markets in investment-grade and high-yield corporate bonds; manages spread risk and inventory; sources scarce bonds. This is the inventory craft — much of the value is simply in finding a bond a client wants when it rarely trades, and pricing the spread risk of holding it.

Repo / financing trader

Funds the firm's inventory, runs the matched book, sources specials and optimises collateral and balance sheet. This desk is the quiet centre of gravity: every position on the floor is financed here, and the leverage ratio makes its balance-sheet discipline everyone's concern.

e-Trading / desk COO / quant

Owns auto-pricing, connectivity, hit-rate analytics and desk automation. This is where technology meets the trading floor, and it is the fastest-growing part of the desk. The quant builds the pricing models; the e-trading lead runs the engines and the venue connectivity; the desk COO runs the business.

Sales and structuring

Bring client flow to the desk, distribute axes and new issues, and package solutions. Sales is the client interface — the relationship, the market colour, the axe distribution — and structuring builds the bespoke trades that go beyond vanilla flow.

Risk, compliance and IT / data

Independent market risk monitors DV01, CS01 and VaR against limits without reporting to the traders. Compliance runs surveillance and the information barriers. IT and data build and run the platform. This is the control and delivery backbone, and the information barrier runs straight through it — risk and compliance must be independent of the desk they oversee.

Team Division Core role Primary technology Maps to stage
Rates trader Markets — rates Make markets in govvies, futures, swaps Streaming pricer, futures gateway Price, hedge
Credit trader Markets — credit Make markets in IG/HY, manage inventory RFQ pricer, inventory / axe tools Price, manage
Repo / financing Markets — financing Fund inventory, run matched book Collateral optimiser, repo platform Finance
e-Trading / quant Markets — e-trading Auto-pricing, connectivity, analytics Pricing engine, FIX gateways Price, execute
Sales & structuring Markets — distribution Client interface, axe & new-issue flow CRM, chat, axe distribution Enquiry
Risk / compliance / IT Control & technology Independent risk, surveillance, platform Risk engine, surveillance, data platform All (continuous)

The technology axis is where most transformation work lands, and it is why this business is worth understanding for technologists as much as traders. Auto-pricing, venue connectivity, composite pricing for illiquid names, and post-trade automation are large, permanent engineering workloads. The desk is increasingly a software product with traders sitting inside it, and the people who build and run that software are as central to the business as the people who quote the prices.

Two actors that are easy to overlook: brokers and data vendors

Interdealer brokers and data vendors sit at the edges of the picture, but the market would not function without them. Interdealer brokers are intermediaries that match dealer-to-dealer flow, often anonymously, and they matter most in the less liquid instruments where a dealer needs to offload risk without revealing its hand. A broker that can quietly find the other side for a large, awkward position — without the whole market learning that a particular dealer is a forced seller — provides exactly the discretion the interdealer market depends on. They are the connective tissue of the sell-side layer.

Data and analytics vendors are the information layer of the market, and their importance grows as the business electronifies. They supply market data, evaluated and composite pricing, indices and transaction-cost analysis — the raw material every pricing engine and every best-execution report consumes. In a market where many bonds do not trade on a given day, the vendor that can supply a credible price for an illiquid name is supplying something close to indispensable. As Part 4 shows, the quality of this information layer is one of the binding constraints on how far the market can automate, which is why data vendors are strategic partners, not mere suppliers.

How sales and trading actually work together

The relationship between sales and trading is the daily rhythm of the desk, and understanding it explains how flow and information move. Sales owns the client relationship; trading owns the risk and the price. A client's enquiry reaches the desk through sales, who knows the account, its history and its likely intention; sales passes it to the trader, who prices it against the book; and the price goes back to the client through sales. This division is deliberate — the trader concentrates on risk and pricing while sales concentrates on the relationship and the market colour — and the handoff between them is where much of the desk's tempo lives.

The axe is the clearest example of sales and trading working as one machine. When a trader wants to do a particular trade — to sell down an ageing position, or to buy back a short as in our running example — the desk publishes an axe, and sales distributes it to the clients most likely to have matching interest. A well-distributed axe brings the desk exactly the flow it wants: the client gets a sharp price, the trader reduces risk, and sales strengthens the relationship. This is why axe distribution is not administrative work but a core part of how a modern desk manages its inventory — turning what the desk needs to do into a service it offers clients, with technology carrying the axe from the trader's book to the right accounts in seconds.

  1. Why does the dealer sit in the middle of every interaction in fixed income?

  2. Contrast how a pension fund and a hedge fund behave as clients. — A pension fund is patient, liability-matched, and price-insensitive at the long end, holding bonds for years; a hedge fund is active, price-sensitive, trades in and out quickly, and can flip positions — it also increasingly supplies liquidity itself.

  3. How does the type of investor affect the price a dealer shows?

  4. Why is the central bank not just another participant?

  5. What is an information barrier, and why is it a technology problem rather than only a policy one?

  6. Name the six desk roles and identify which one is independent of the traders and why.

3

  1. Why does the dealer sit in the middle of every fixed income interaction?

  2. How does client type affect the price shown?

  3. Why is the central bank not just another participant?

  4. Which client is patient, liability-matched and price-insensitive at the long end?

4Systems, Data and the Technology That Carries the Trade

The market only works because of the plumbing. This part opens up the systems, follows the data front to back, and shows why fixed income is a uniquely hard data problem.

A fixed income trade is, from the technology side, a data object that is born at pricing and travels through a dozen systems before it comes to rest in a regulatory report. If you understand that journey — what each system does, what data it creates, and how an error in one reaches all the others — you understand the operating model of the whole business. This part starts from the clean question of what functions any deal needs, names the vendors that fill them, follows one trade front to back, opens up the data model, and covers the messaging, identifiers, resilience and regulation that hold it all together. It is written to be useful to a technologist building the platform as much as to a trader using it.

Start with functions, not vendors

Before naming any system, ask what functions a fixed income trade actually needs — the answer is the same for every bank. Strip away the brand names and every trade requires the same handful of capabilities in sequence: something to price and quote it, something to capture and manage the order, something to book it and measure the risk, something to clear and settle it, something to finance it, and something to keep the books and file the reports. Around that chain sit cross-cutting capabilities that touch every stage. This functional view is the right map because vendors come and go but functions do not.

Figure 12 The functional architecture — six blocks from quote to report, over a band of cross-cutting capabilities
Functional block What it must do Cross-cutting capabilities
Pricing & quoting Build curves and spreads, derive a fair mid, auto-respond to RFQs Market data & composite pricing
Execution & OMS Capture the trade from voice and every platform Venue & FIX connectivity
Risk & booking Book instantly; compute positions, DV01/CS01, P&L Reference & instrument data
Clearing & settlement Confirm, clear via CCP where applicable, settle DVP at CSD Compliance & surveillance
Financing & repo Fund inventory, optimise collateral, run the matched book Straight-through processing
Books & reporting Keep books and records; file MiFIR / SFTR / EMIR reports Operational resilience

The vendor landscape — the names you will meet, and buy versus build

Once you know the functions, the vendor map is easy to read: banks buy the commoditised layers and build the layers that are their edge. Venues, market data and post-trade utilities are bought — there is no advantage in reinventing a trading platform or a settlement pipe. Pricing engines, risk systems and the decision logic that skews a quote are built or heavily customised, because that is where competitive advantage lives. The buy-versus-build line is therefore a map of where a bank thinks it can win.

Layer Representative names Buy or build Why
Multi-dealer venue (rates & credit) Tradeweb Buy Shared market infrastructure; no edge in rebuilding
Credit venue & all-to-all MarketAxess (Open Trading) Buy Network effects belong to the platform
Terminal, chat, RFQ & data Bloomberg Buy Ubiquitous; where flow originates
Pricing, curves & auto-quoting In-house + vendor engines Build The desk's edge — proprietary skew and models
Risk & booking In-house / vendor risk system Build / customise Firm-specific limits, P&L and controls
Interdealer & post-trade data IDBs; TRACE (US) / EU tape Buy Reference data informs pricing and TCA
Clearing & settlement LCH, Eurex, CSDs, utilities Buy Regulated market infrastructure

The platforms are worth knowing by name because they define the electronic market. Tradeweb is a leading dealer-to-client venue across government bonds, rates, credit and ETFs, offering RFQ, streaming and portfolio trading. MarketAxess is the dominant corporate-bond platform and the pioneer of all-to-all trading through its Open Trading model. Bloomberg is the ubiquitous terminal, chat, RFQ and market-data layer where a huge share of flow originates. Underneath them, in-house and vendor pricing and risk engines build the curves, price the bonds and manage the book, and the post-trade utilities — CCPs, CSDs and reporting services — clear, settle and report.

Figure 13 Buy the commoditised layers; build the layers that are the desk's edge

The buy-versus-build line is not fixed once and forgotten; it moves as the market commoditises. A capability that was once a proprietary edge — a basic RFQ pricer, say — becomes a commodity once every bank and vendor has one, and the line shifts up to the next frontier, which today is composite pricing for illiquid names and the analytics that decide which flow to win. Reading a bank's buy-versus-build choices is therefore a way of reading where it thinks the competitive frontier now sits: it builds where it believes it can still be better than the market, and buys everything below that line. The steady direction of travel is that more of the stack commoditises over time, pushing the edge into data and into the decision logic that data feeds.

Following one trade front to back

Trace a single trade through the systems and the operating model becomes concrete. An RFQ hits the pricing engine, which generates a quote and, for liquid flow, auto-responds. The order and execution management system captures the trade from voice or platform. The trade books instantly into the risk system, where the position, DV01, CS01 and P&L update. Confirmation follows, then CCP clearing where applicable, then settlement through the CSD. Finally the position is financed in repo, and the trade feeds books-and-records and regulatory reporting. Market data feeds pricing throughout; risk and compliance run through the whole chain; and straight-through processing means no manual re-keying from front to back.

Figure 14 One trade, one identifier — an error in the price at the front reaches every downstream system
Stage System Data object created
Pricing & quoting Pricing engine Quote + fair mid (curve, spread, axe)
Execution OMS / EMS Trade capture (price, size, counterparty, timestamps)
Risk & booking Risk system Position, DV01/CS01, P&L, limit usage
Clearing CCP interface Cleared trade, novated to the CCP
Settlement CSD interface DVP instruction, settled record
Financing Repo / collateral system Financing trade, collateral allocation
Reporting Regulatory reporting engine MiFIR / SFTR / EMIR reports + audit trail

The anatomy of the atom — the trade and order data model

The whole back end is built from one small object: the trade record. Understanding its fields is understanding the plumbing. When a trade is captured, a handful of fields are set, and each one drives something specific downstream. Get the fields right and the trade flows straight through; get one wrong and it breaks somewhere far away.

Field What it holds What it drives downstream
Instrument identifier (ISIN) Which exact bond Pricing, settlement routing, reporting
Direction & size Buy/sell, notional Position, DV01/CS01, settlement amount
Price / yield / spread The agreed level P&L, TCA benchmark, reporting
Counterparty (LEI) Who the client is Credit limits, clearing, reporting
Timestamps When quoted, executed, booked Best-execution evidence, surveillance
Venue / channel Where it was done Best-execution policy, transparency
Trader / desk / book Who owns the risk Risk aggregation, P&L attribution

Deduplication is a genuinely hard, mission-critical problem, not a clerical one. A single trade can arrive through more than one path — a platform feed and a voice ticket, a primary and a backup connection — and if the systems cannot tell that two records are the same trade, the bank books it twice, doubling a phantom position and misstating its risk and its regulatory reports. Reliably recognising that two messages refer to one economic event, across venues and channels and under time pressure, is one of the quiet, hard problems of trade processing. It is also exactly the kind of problem where clean identifiers and disciplined data engineering earn their keep.

Identifiers — the three codes that make the plumbing work

Three identifiers do the heavy lifting of connecting systems, and confusing them is a common source of error. Each answers a different question — which security, which legal entity, and how to map an instrument across vendors — and each "bites" at a different point in the chain.

Identifier What it names Where it bites
ISIN The specific security (this exact bond line) Pricing, settlement, transaction reporting
LEI The legal entity (the counterparty or issuer) Credit, clearing, all regulatory reporting
FIGI A vendor-neutral instrument mapping Reconciling instrument data across systems and vendors

The ISIN (International Securities Identification Number) names the exact bond — a given issuer's five-year 3% line, distinct from its ten-year line. The LEI (Legal Entity Identifier) names the legal entity on the other side, which every regulatory report demands. The FIGI (Financial Instrument Global Identifier) provides a vendor-neutral way to map the same instrument across different data providers and systems, which matters enormously when a bank runs dozens of systems that each have their own internal code for the same bond. Together they are the coordinate system of the whole market.

Integration patterns — how the systems actually talk

Three patterns connect the systems, and the migration between them is one of the defining technology stories of the desk. The oldest is batch: systems exchange files at fixed times, typically overnight. It is simple and robust but slow, and its latency means positions and reports can be hours out of date — and it forces constant reconciliation, because two systems only agree after the next batch runs. The modernisation is the API: one system asks another for data on demand and gets an immediate answer, so information is current when it is needed. The newest is event streaming — a Kafka-style continuous push where every change is broadcast as it happens and every interested system updates in real time.

Immutable audit and the decision engine

Some records must never be changed after the fact, and some logic must never be bought off the shelf. Compliance-critical records — who quoted what, when a trade was executed, what a surveillance alert found — must be append-only: you can add to them but never silently edit or delete them, because their entire value is that they are a trustworthy history a regulator can rely on. Governance and audit platforms enforce this immutability. At the other end, the decision engine that turns market data and inventory into a skewed quote is always proprietary — it is the desk's edge, the codification of its pricing judgement, and no bank would outsource the thing that makes it money. The pattern is worth remembering: buy the immutable record-keeping, build the decision logic.

Messaging protocols — two worlds, one trade

The front office and the back office speak different languages, and understanding the split explains the shape of the whole technology stack. In the front office, the language is FIX — the Financial Information eXchange protocol. FIX carries RFQs, quotes and executions between clients, venues and the bank in real time; it is fast, standardised, and it is how an electronic trade enters the primary process. Once the trade is done, it crosses into the post-trade world, whose language is SWIFT — the messaging network of settlement instructions and confirmations. Post-trade messaging is migrating from the older SWIFT MT format to the richer ISO 20022 (MX) standard, which carries more structured data and supports far more automation.

Figure 15 Front office speaks FIX; post-trade speaks SWIFT MT then ISO 20022 — two messaging worlds, one trade
Layer Standard What it carries Timescale
Front office FIX RFQs, quotes, orders, executions Real time (sub-second)
Post-trade SWIFT MT → ISO 20022 (MX) Confirmations, settlement instructions Minutes to T+2

Operational resilience — because the critical moment has no rewind

Some moments in a trading day cannot be repeated, and the systems must be built so those moments never fail. A pricing engine that goes down mid-session, a venue link that drops during a volatility spike, a settlement pipe that jams on a large delivery — each is a moment with no second chance, where a few minutes of outage means lost trades, broken quotes, or failed settlements with real financial penalties. Resilience is the discipline of engineering so that these moments survive any single failure.

Concept What it guarantees Why it matters on the critical day
Redundancy No single point of failure — components are duplicated A dead server does not take the desk offline
Automatic failover Traffic switches to the backup with no human action Recovery in seconds, not the minutes a human would take
Disaster recovery A separate site can take over the whole system A data-centre loss does not stop the business
Business continuity People and processes keep running through disruption The desk keeps trading when the building or region is hit

Connectivity and market data — the fuel and the wiring

None of the electronic market works without connectivity to venues and a constant stream of clean market data, and in fixed income the data problem is uniquely hard. The desk must be wired to every relevant platform — each with its own onboarding, rules and API — and it must ingest real-time prices, curves and spreads from venues, vendors and evaluated-pricing services. The distinctive difficulty is that many bonds do not trade on a given day, so there is no live price to ingest; the bank must buy or build composite and evaluated prices to mark and quote illiquid names. Building robust venue connectivity, composite pricing, and full straight-through processing is therefore a large, permanent technology workload — and it is exactly where the transformation agenda in Part 5 concentrates.

Composite and evaluated pricing — inventing a price where none exists

The hardest part of the market-data problem deserves naming on its own. When a bond has not traded for days, there is no observed price to feed the engine — so the bank must construct one. A composite price blends the available quotes and trades from many venues and dealers into a single consensus mark; an evaluated price goes further, modelling a value from the government curve, comparable bonds, and the issuer's other lines. These constructed prices are what let a desk mark its book, quote an illiquid name, and evidence best execution where no benchmark otherwise exists. They are also inherently uncertain, which is why a model price on an illiquid bond must always be treated with the humility Part 5 insists on. The coming EU consolidated tape for bonds — a single feed of post-trade prices — is significant precisely because it promises to give this constructed-price problem a firmer factual backbone.

Speed as a feature — latency, co-location and the tick-to-trade path

In the liquid, streaming corners of the market, speed is not a nicety but a feature that decides who wins the trade, and building for it is a distinct engineering discipline. When a dealer streams live prices that clients click to deal, or quotes govvies and futures where the market moves continuously, the time between a market-data update arriving and a fresh quote going out — the tick-to-trade path — determines whether the desk is quoting a stale price that loses money or a current one that wins. Every layer the trade passes through adds latency, and shaving microseconds off each is a real, funded objective.

Figure 16 The tick-to-trade path — every layer adds latency, and co-location shrinks the round-trip

The path runs through a stack, and the physics of it matters. A market-data feed delivers prices in; the pricing engine computes a fair mid and skew; a pre-trade risk-and-limits layer checks the quote is safe; a venue gateway speaking FIX sends it out; and the venue matches and acknowledges. The single biggest lever on the round-trip is distance: signals travel at a finite speed, so a bank places its servers physically next to the venue's — co-location — cutting the network path from kilometres to metres and the round-trip from milliseconds toward microseconds. This is why low latency is as much a data-centre and networking problem as a software one, and why it barely matters for illiquid credit (where a bond may not trade for days) but is decisive for liquid rates. The lesson is to build for latency only where speed actually changes the economics — a specific case of the transformation filter in Part 5.

Controlling the machine — automated-trading controls and kill switches

Once pricing and execution are automated, the danger shifts from a trader's mistake to a machine's, and the controls have to be automated too. An engine that auto-quotes thousands of times a second can, if something goes wrong — a bad data feed, a mispriced curve, a runaway loop — do damage far faster than any human. So automated trading is wrapped in a layer of hard, pre-trade controls that sit in the tick-to-trade path itself, checking every machine-generated order before it leaves the building. This is not optional good practice; in Europe the rules on algorithmic trading (the MiFID II regime and its technical standards) require it explicitly.

The controls are concrete and worth naming. Pre-trade risk checks validate every order against price and size limits, credit lines and position limits before it is sent — an order outside the bounds is blocked, not sent and cancelled. Throttles cap how many orders the engine can fire in a given interval, so a runaway loop cannot flood the market. Price collars reject quotes that are implausibly far from the fair mid, catching a bad data feed before it prices a trade. And above all sits the kill switch: a single control that instantly pulls every quote and halts the engine, so a human can stop the machine the moment its behaviour looks wrong. The whole design assumes the automation will sometimes misbehave and makes that misbehaviour survivable.

The golden source — reference data, the security master and break management

Every system in the chain has to agree on what a bond *is*, and keeping that agreement is a permanent, underrated discipline. A trade record carries an ISIN, but to price, risk-manage, settle and report it, dozens of systems each need the bond's full details — coupon, maturity, day-count, currency, seniority, ratings, the issuer's LEI. If two systems disagree about any of these, the trade breaks somewhere downstream. The answer is a security master: a single, authoritative reference-data store — a golden source — that every system draws the instrument's details from, so they cannot drift apart. Maintaining it, and onboarding every new bond into it cleanly and promptly, is a genuine operational workload, because a bond that is not correctly set up in the master cannot be traded properly no matter how good the pricing engine is.

When systems do disagree, the difference is a break, and resolving breaks is a whole workflow. A break is a mismatch — the risk system and the books show different positions, or a confirmation does not match the booked trade, or a settlement instruction is rejected. Reconciliation is the process of finding breaks by comparing systems, and break management is the workflow of investigating and fixing each one before it causes a failed settlement or a misstated report. This is exactly the class of work that event streaming, from Part 4's integration patterns, aims to eliminate at the source: if every system consumes the same events in the same order, there is nothing to reconcile. Until that migration is complete, reconciliation and break management remain a large, unglamorous part of running the platform — and a prime target for automation.

Five regulatory regimes, read as system specifications

Regulation shapes every system in the chain, and the productive way to read each rule is as a specification the platform must meet. These are not legal footnotes; each one dictates what data the systems must capture, how they must behave, and what they must be able to prove. The five below are the European framework, and each translates directly into a technical requirement.

Regime What it demands System implication
MiFID II / MiFIR Pre/post-trade transparency, best execution, the SI regime Capture full timestamps, evidence execution, publish trades, transaction-report
MAR (Market Abuse Regulation) Bans insider dealing and manipulation; governs conduct Trade & communications surveillance, restricted lists, alerting
SFTR Two-sided reporting of every repo / securities-financing trade Full financing data capture and matched reporting
EMIR Central clearing and margining of standardised swaps; derivative reporting Clearing connectivity, margin calculation, trade reporting
Basel / FRTB / leverage ratio Capital and leverage against trading risk Risk-weighted capital and balance-sheet measurement per position

Read this way, the compliance burden becomes an engineering brief. MiFID II says: capture every timestamp, evidence best execution, publish and report trades. MAR says: run surveillance over trades and communications, maintain restricted lists, alert on abuse. SFTR says: report every repo, from both sides, with matching fields. EMIR says: clear the standardised swaps, calculate the margin, report the derivatives. Basel and FRTB say: measure the capital and balance-sheet cost of every position. Each regime is a set of data and behaviour requirements the platform must satisfy — which is why regulatory reporting is one of the heaviest, most integration-hungry parts of the whole build, and why consolidating MiFIR, SFTR and EMIR reporting is a recurring transformation theme.

  1. Why is it better to map the architecture by function before naming vendors?

  2. State the cascade principle and why it makes data quality at capture so important.

  3. What are the three key identifiers and what does each name?

  4. Why is deduplication described as mission-critical?

  5. What is the difference between FIX and SWIFT, and where does each operate?

  6. Why is redundancy alone insufficient for resilience?

  7. Why is migrating from batch to event streaming more than a speed upgrade?

  8. What is a kill switch, and why does automated trading need one?

4

  1. What does each identifier name?

  2. Why is deduplication mission-critical?

  3. Why map the architecture by function before vendors?

  4. Why does the cascade principle make data capture so important?

5Risks, Failures and the Transformation Agenda

Every business is best understood by how it goes wrong. This part maps the risks, tells three true stories of failure, and separates real transformation from noise.

The previous parts built the business as it works. This one studies it under stress. A fixed income desk runs a specific set of risks, and each has an owner and a defence; when the defences fail, the failures fall into a small number of recognisable patterns, and three real cases illustrate them. The part then turns forward: to the technology trends genuinely reshaping the desk, to the honest role of artificial intelligence, and finally to the six ideas worth keeping from the whole module. The register throughout is the same one a risk committee would use — precise about what can go wrong and clear-eyed about what actually helps.

The six risks a dealer runs

A fixed income desk faces six distinct risks, and the value of naming them exactly is that each has a different owner and a different defence. Lumping them together as "market risk" hides the fact that they are controlled in different places by different people with different tools. The taxonomy below is deliberately MECE — mutually exclusive, collectively exhaustive — so that every loss a desk can suffer maps to exactly one category.

Figure 17 The six trading risks — a map of controllability, each with an owner and a defence
Risk type What it is Who owns it Main defence
Market (price / rate) Prices move against the position while it is held The trader DV01 limits, futures/swap hedges
Spread (credit) The credit spread widens on held bonds The credit trader CS01 limits, CDS/index/ETF hedges
Inventory / liquidity A position cannot be exited at a fair price The desk / trader Concentration & age limits, recycling
Funding (repo) Financing cost rises or dries up The financing desk Diversified funding, term matching
Counterparty / settlement The other side fails to pay or deliver Risk & operations CCP clearing, DVP, credit limits
Conduct / operational Misconduct, error, or a broken process Compliance & COO Surveillance, controls, STP, audit

Telling a good desk from a lucky one

A desk can look profitable for the wrong reasons, so it is judged on quality metrics, not just P&L. A trader who makes money by quietly running oversized directional risk is not a good market-maker; they are a lucky gambler who will eventually be unlucky. The metrics below separate durable, franchise-building performance from the appearance of it.

Metric What it measures What "good" looks like
Hit-rate (RFQ win rate) Share of quoted RFQs won High enough to matter, not so high the desk is underpricing
Spread capture Realised bid-offer per trade Positive after hedge and funding cost
Inventory turnover How fast positions are recycled Fast — young, moving inventory
Risk-adjusted P&L Return per unit of VaR / balance sheet High — earning on scarce resources
Client quality & breadth Depth and diversity of the client base Broad, sticky, real-money weighted

Three true stories of how it goes wrong

Fixed income failures are not random — they cluster into three modes, and one famous case illustrates each. The first mode is bad risk management: a position grows too large and too concentrated, and the controls that should have caught it fail. The second is a bad market: liquidity vanishes exactly when it is needed, and even a sound position cannot be exited. The third is a bad model with too much leverage: a relative-value bet that looked riskless proves anything but. The three cases below are real, and together they are the failure modes of the whole business.

Case one — the London Whale (JPMorgan, 2012): concentration and control failure

In 2012, a trader in JPMorgan's Chief Investment Office in London — nicknamed the "London Whale" for the size of his positions — built an enormous, concentrated bet in credit derivatives (CDX index products) that was meant to hedge the bank but became a vast directional position in its own right. As the market moved against it, the position was too large to exit without moving the price further against itself, and internal risk controls and valuations had failed to flag the danger in time. The loss ran to roughly six billion US dollars. This is the concentration-and-control failure mode: the risk was a credit-spread position (CS01) that grew beyond what the market could absorb, and the defences — limits, independent valuation, surveillance — did not bite. It maps to the hedge-and-manage step, and its lesson is that a hedge that grows into a concentrated directional position is no longer a hedge, and that risk controls are only worth anything if they are independent and enforced.

Case two — the UK gilt / LDI crisis (2022): liquidity vanishes when it is needed most

In late September 2022, a UK fiscal announcement triggered a violent sell-off in gilts — UK government bonds. The move was amplified by pension funds running liability-driven investment (LDI) strategies, which faced collateral calls as yields spiked and were forced to sell gilts to raise cash, driving yields higher still in a self-reinforcing spiral. Dealer balance sheets, constrained by post-crisis capital rules, could not absorb the wave of forced selling, and the market seized until the Bank of England intervened with emergency bond purchases. This is the bad-market failure mode: liquidity was abundant in calm conditions and evaporated precisely when it was needed, exactly the state-dependence described in Part 1. It maps to the whole market-making function, and its lesson is that liquidity is a fair-weather promise — a dealer's ability to warehouse risk shrinks just when clients most need it, and leverage hidden elsewhere in the system (here, in pension collateral) can turn a price move into a spiral.

Case three — Long-Term Capital Management (1998): the relative-value model that broke

Long-Term Capital Management was a hedge fund staffed by star traders and Nobel laureates that ran exactly the kind of relative-value trades described in Part 2 — betting that small mispricings between related bonds would converge — but at enormous leverage. When Russia defaulted in 1998, markets did the opposite of what the models assumed: spreads that were "supposed" to converge diverged violently, and because the positions were so leveraged, the losses threatened the whole financial system, forcing a coordinated rescue. This is the bad-model, too-much-leverage failure mode: relative-value trades are low-risk per unit only if the assumed relationships hold and the leverage is survivable, and both assumptions failed at once. It maps directly to the relative-value trade in Part 2, and its lesson is that a hedged, "market-neutral" position is only neutral within the model's assumptions — extreme events break correlations, and leverage turns a broken assumption into an existential loss.

Case Failure mode Step it maps to The lesson
London Whale (2012) Concentration & control failure Hedge & manage A hedge that becomes a concentrated bet is not a hedge; controls must be independent and enforced
UK gilt / LDI (2022) Liquidity evaporation / bad market Market-making Liquidity is state-dependent; it vanishes when most needed, and hidden leverage amplifies
LTCM (1998) Bad model + excess leverage Relative-value trade "Market-neutral" holds only within the model; extreme events break correlations

The pain points every fixed income business is trying to solve

Before the transformation agenda, it helps to name the problems it is trying to fix — and almost every one is a data or a balance-sheet problem in disguise. These are the recurring frustrations of running a modern fixed income business, and they are remarkably consistent across banks.

Pain point What it is Underneath, it is a...
Liquidity fragmentation Flow split across many venues, protocols and thousands of bond lines Data problem — no single view of the market
Balance-sheet & inventory cost Capital and leverage rules make holding inventory expensive Balance-sheet problem — scarce, costly capacity
Data quality & coverage Many bonds lack a live price; reference data and timestamps are messy Data problem — the raw material is patchy
Pricing illiquid bonds Auto-quoting breaks where there is no recent trade Data problem — modelling a price from thin signals
Regulatory reporting load MiFIR, SFTR, EMIR and CSDR each demand different data, reconciled by hand Data problem — fragmented, duplicated reporting
Legacy & fragmented systems Rates, credit and repo run on separate ageing platforms with no shared data Data problem — no straight-through processing

The common thread is the whole transformation thesis in one line: fix the data and use the scarce balance sheet more cleverly. Five of the six pain points above are data problems, and the sixth is a balance-sheet problem. That is not a coincidence — it is why the transformation agenda is dominated by better data (composite pricing, reference data, transparency) and by ways to move risk without warehousing it (electronification, all-to-all). Everything in the rest of this part follows from that diagnosis.

The next frontier — tokenisation and distributed ledgers

A quieter but potentially deeper shift than AI is the move to represent bonds themselves as digital tokens on a distributed ledger, and it is worth understanding honestly — neither dismissed nor hyped. A tokenised bond is a bond whose ownership lives as a record on a shared, cryptographically secured ledger (a blockchain or similar distributed-ledger technology, DLT) rather than in the traditional chain of custodians and a central securities depository. Several governments, development banks and large issuers have already issued bonds this way, and major market infrastructures are building tokenised-settlement platforms. It is no longer a thought experiment.

The reason it could matter reaches straight into the plumbing this module has described. Recall that settlement takes one or two days, runs through a CSD, and closes with delivery-versus-payment — and that much of Part 4's difficulty was reconciliation between systems that hold their own copy of the truth. A distributed ledger attacks exactly those frictions: if the bond and the cash both live as tokens on a shared ledger, ownership can transfer atomically — bond and cash moving together, instantly, with no separate reconciliation because there is only one record. In principle that means near-instant settlement (compressing T+2 toward T+0), a single golden source by construction, and programmable instruments that can automate coupon payments and corporate actions. Smart contracts could even carry repo and collateral logic directly.

Generative AI on the desk — where it earns its place, and where it does not

Artificial intelligence, including large language models, has real and growing uses in fixed income trading — and specific risks that must be governed as tightly as the value is pursued. The honest position is neither hype nor dismissal. AI helps most where the problem is data-shaped and the stakes of a single output are bounded; it is dangerous where a confident-looking answer on sparse data can drive a real loss.

The genuine uses cluster in a few areas. Liquidity prediction: models estimate which bonds can trade, in what size, and at what level, guiding pricing and axes in a market where that information is otherwise guesswork. Pricing illiquid bonds: machine-learning fills the missing price for names that rarely trade, inferring it from comparables and signals — directly addressing the hardest data problem in the business. Smart RFQ and hedging: analytics decide which RFQs are worth winning, at what price, and how to hedge the residual. Data extraction and operations: language models read documents, reconcile breaks, and automate confirmations and reporting — unglamorous work where automation pays immediately. And client and flow analytics: predicting client needs and targeting axes to the right accounts.

Use of AI / GenAI The gain The governing risk
Liquidity prediction Better pricing and axing in a data-poor market Overconfidence on sparse data
Auto-pricing illiquid bonds A usable price where none is quoted A precise-looking price that is simply wrong
Smart RFQ & hedge selection Higher hit-rate, cheaper hedging Model risk; needs limits and controls
Document & ops automation (LLMs) Faster confirmations, fewer breaks Accuracy, and confidentiality of inputs
Client & flow analytics Right axe to the right account Data protection; MNPI and privacy

The caveats are not disclaimers; they are the design constraints. Data is the bottleneck — a model is only as good as the sparse, patchy bond data it learns from, and garbage in is garbage out. Judgement stays human — pricing a distressed name or a stressed market still needs a trader's experience, because that is exactly where the historical data is thinnest and least reliable. There is a real risk of false confidence — a model price on an illiquid bond can look precise to three decimals and be badly wrong, so controls and limits are essential. And deployment must be governed — model-risk rules, explainability requirements, and clear accountability mean AI on the desk is human-in-the-loop by design, not autonomous. There are also confidentiality risks specific to this business: feeding material non-public information or client data into a model is a serious control failure, so what goes into a model is governed as strictly as what comes out.

How the governance actually works

Model-risk governance is not a slogan; it is a concrete process, and it is worth knowing its shape. A model used to price bonds or decide RFQs is independently validated before it goes live — a team separate from the one that built it checks that it does what it claims and fails safely. It is documented so that its logic can be explained to a regulator or a risk committee, which is what "explainability" means in practice: not that every parameter is intuitive, but that the model's behaviour can be accounted for. It runs within limits, so that even a wrong output cannot cause an unbounded loss. And a named person is accountable for it. Applied to artificial intelligence, these same four disciplines — validation, documentation, limits and accountability — are exactly what keep a machine-learning price on an illiquid bond from being trusted blindly. The human in the loop is where the accountability lands.

5

  1. Which risk is defended by DV01 limits and futures or swap hedges?

  2. What does the London Whale case teach?

  3. What failure mode does the 2022 UK gilt/LDI crisis illustrate?

  4. A trader makes money by quietly running oversized directional risk. How should the desk be judged?

The six ideas that matter most

Idea 1 — There is no exchange, so the dealer is the market

Everything follows from this. In a market with millions of rarely-traded bonds and no central order book, the dealer's willingness to quote a two-way price and warehouse the risk is the liquidity. Immediacy is the product; the bid-offer spread is the price of it; and balance sheet, not headcount, is the binding constraint on how much can be provided.

Idea 2 — Data cascades, so an error is never contained

A trade is a data object born at pricing and travelling through a dozen systems to a regulatory report. Every downstream record trusts the one before it, so an error in the price at the front reaches risk, settlement, reporting and the books. The trade record is the atom; deduplication is a hard problem; and data quality at capture is the most leveraged control in the whole architecture.

Idea 3 — Pricing and hedging are where the money and the risk are made

The lifecycle's danger concentrates in the middle, where the dealer prices the trade and takes it onto its book. A two-way price is built in four moves — fair mid, spread, skew for the axe, hedge and funding cost — and a corporate bond's two risks (rate and spread, DV01 and CS01) must be measured and separated. The client's type is a hard input to the price, not soft relationship-building.

Idea 4 — Compliance and resilience are core architecture, not afterthoughts

The information barrier, the surveillance, the immutable audit trail, the automatic failover — these are not bolted on after the fact; they are load-bearing parts of the platform. Regulation is best read as a system specification, and the moments that cannot be repeated — a pricing night, a settlement of size — demand resilience engineered for the stressed day, not the calm one.

Idea 5 — Balance sheet is scarce, and that quietly shapes every decision

Post-crisis rules — Basel III, the leverage ratio, FRTB — made dealers safer but made their balance sheets scarcer and more expensive. Banks now price balance-sheet cost into every trade, prioritise high-return flow, and lean on electronification and all-to-all to move risk without warehousing it. League tables reward volume; the real measure is return on the scarce resource.

Idea 6 — Real transformation moves cost or risk — including AI, with governance

Electronification, portfolio trading, all-to-all, straight-through processing and AI are reshaping the desk, but only where they change the economics or the risk of the business. A digital veneer over a manual process changes neither. AI's honest value is efficiency and better signals in a data-poor market, governed human-in-the-loop; better data, not a bigger model, is the real unlock.

  1. Name the six risks a fixed income desk runs and give the main defence for two of them.

  2. Why is traded volume a poor measure of a desk's quality, and what is better?

  3. Match each of the three cases to its failure mode.

  4. What is the filter that separates real transformation from decoration?

  5. Where does generative AI genuinely help on a fixed income desk, and why must it stay human-in-the-loop?

Process Cheat Sheet — The Whole Desk on One Page

A single reference for the entire fixed income trade flow, from a client enquiry to a settled, financed position.

The seven-step lifecycle below is the spine of the business. Read it top to bottom and you have followed a trade from the client's first request to its settled, financed resting place on the book. The illustrative figures are orders of magnitude, not precise facts.

Stage Timing Core activity Key output Lead & systems
1. Enquiry Seconds Client asks for a price (RFQ or voice) A request on a specific bond Sales / platform; RFQ venue
2. Price the risk ms – min Fair mid, spread, skew, hedge & funding cost A bid / offer Trader / pricing engine
3. Quote & negotiate Seconds Show a two-way price; best price wins A firm price Trader / auto-quoter
4. Execute & book Instant Trade agreed, captured, risk updates A booked position OMS/EMS; risk system
5. Hedge & manage Min – days Hedge rate risk, isolate spread, recycle A hedged position in limits Trader; futures/swaps
6. Confirm & clear T+0 Confirm economics; CCP clears where applicable An agreed, cleared trade Operations; CCP
7. Settle & finance T+1/T+2, ongoing DVP exchange; fund position in repo A settled, financed position Operations; repo desk

The four-move price is the recipe every quote follows: find a fair mid, add a spread for size and illiquidity, skew for the desk's axe, and factor in the hedge and funding cost. The two-risk model is the daily discipline: a corporate bond carries rate risk (DV01) and spread risk (CS01), measured and hedged separately. The three revenue engines are bid-offer spread, carry, and financing, sitting on top of a hedged flow franchise. And the two messaging worlds are FIX in the front office and SWIFT MT (moving to ISO 20022) in post-trade.

Metric Rule of thumb (illustrative)
Settlement cycle T+1 or T+2, delivery-versus-payment
Rates electronic share ~60–70% (govvies, futures, vanilla swaps)
Credit electronic share ~40–50% and rising (IG ahead of HY)
Primary rate risk measure DV01 — P&L per 1bp move in yields
Primary credit risk measure CS01 — P&L per 1bp move in spread
Carry Bond yield minus repo funding cost, accrued daily
Firm-wide risk limit VaR, backed by scenario and stress tests
Repo stress dates Quarter-end / year-end (window dressing)
Binding capital constraint Leverage ratio (bites on low-margin repo & govvies)
Key European regulations MiFID II/MiFIR, MAR, SFTR, EMIR, Basel/FRTB, CSDR

The whole desk in one sentence: because there is no exchange, a dealer quotes a two-way price on a bond, takes the risk onto its own balance sheet, hedges away what it does not want and finances what is left, and moves the trade safely through clearing and settlement — getting paid the bid-offer spread for providing the immediacy no exchange can.