What MarkitWire actually does (and what it doesn't)

The morning starts at 07:45 with MarkitWire. A buy-side ops head, the desk lead at L&G's rates book, opens the platform and sees 73 trades waiting for affirmation. Most are routine: 5-year SOFR-Compound IRS, $10m notional, the kind that books and clears every day at 10 Paternoster Square. The platform shows the dealer's view next to her trader's instructions. She clicks affirm. Repeats.

Sixty-eight clear in fifteen minutes.

Five don't. One has a 2 basis point spread the dealer didn't tell her about. Another has the wrong roll convention. A third was booked with an effective date one business day off because the SEF's UTI assignment rule disagreed with her OMS's. A fourth had been amended after lunch yesterday and she missed the email. The fifth, well. She'll spend the rest of her morning on the fifth.

That's MarkitWire. A shared affirmation blotter that says: here is what I think we agreed; do you agree?

It's a workflow tool, not a matching engine.

What it actually does

MarkitWire was Markit's first product. Lance Uggla launched it in 2001 from a converted dairy in St Albans, before Markit was the data company that bought up CDS pricing services and got to a $44 billion valuation when S&P Global swallowed the lot in March 2022. The original idea was simple: get the buy-side and sell-side off Bloomberg messages and faxes for trade confirmation. Replace them with a structured workflow that two parties could both see at the same time.

The platform did exactly that. It still does. You book a trade in your OMS, the OMS pushes a message to MarkitWire, the counterparty's OMS does the same, and MarkitWire shows you both views in a single screen. You affirm or reject. If you affirm, the trade gets stamped, an audit trail records the timestamps, and the regulatory affirmation clock under EMIR Refit RTS Article 9 is satisfied. The platform processes millions of confirmations a year across rates, credit, and equity OTC.

That's a real piece of work. It replaced fax. It standardised the document templates the legal team had to argue about. It sat alongside ISDA's master agreements as the operational layer underneath them. By 2010 the major dealers and the bigger asset managers were on it. By 2015 it was, for IRS, the network. If you weren't on MarkitWire you couldn't trade with the people who were, and the cost of not being on it dwarfed the cost of being on it.

In September 2021 CME and IHS Markit folded MarkitWire, TriOptima, Traiana, and a couple of other workflow assets into a JV called OSTTRA. Day one, OSTTRA had something like 80% of the post-trade workflow market for OTC rates. The S&P merger followed six months later. Today, when an L&G ops head clicks affirm, the bytes flow through infrastructure that's part-S&P, part-CME, branded OSTTRA, running the platform Lance Uggla built in 2001.

It's a workflow tool, not a matching engine. The distinction matters.

What it doesn't do

MarkitWire matches confirmations, not terms.

Here is what that means. When two parties send their views of the same trade to MarkitWire, the platform compares them at the document level: same counterparty, same trade ID, same broad fields, same end-date. If the documents agree, it's a match. If they don't, MarkitWire flags a discrepancy and shows the two documents side by side. The ops team has to read them and figure out which field differs.

That's the bit that takes 30 minutes when it should take 30 seconds.

A trade has, depending on how you count, between 40 and 120 economic terms. Notional. Fixed rate. Day count. Payment frequency. Reset frequency. Fixing offset. Business-day convention. Calendars. Roll convention. Stub handling. Each one of those, in a real OTC IRS, can be the field that differs. MarkitWire won't tell you which one. It tells you the trades don't match. Then it puts the two confirmations on screen and trusts you to scan them.

So you scan. And scan. The reason a typical break takes 30 minutes is that "compare two trade confirmations field by field, decide which one is right, get the dealer on the phone, agree, amend in the OMS, re-submit" is a 30-minute job. Sometimes it's a 30-minute job because the answer is obvious once you find the field. Sometimes it's a 30-minute job because someone on the dealer side has gone to lunch.

There's a structural reason for the design. MarkitWire was built around legal confirmations because in 2001 the problem was that confirmations were paper. The platform digitised the document flow. It didn't reduce the document to its economic content and reason about that. It couldn't, really. The canonical-form work that ISDA's CDM eventually started in 2018 wasn't on the table.

So the design baked in a particular assumption: the unit of work is the confirm. You compare confirms. A break is a confirm-level event.

The unit of work, for ops, is the field. A break is a field-level event.

Why it persists

If MarkitWire's wrong abstraction is so obvious, why hasn't anyone replaced it? Three reasons, and only one of them is technical.

The first is the network. Once a buy-side desk uses MarkitWire to talk to UBS and JPMorgan and Goldman, switching means coordinating all three sell-side ops floors plus the buy-side's own. Nobody moves first. The cost of being the first off-network is being unable to confirm trades with the rest of the street, and the value of being on a slightly better platform doesn't clear that bar for a single counterparty pair.

The second is regulatory. EMIR Refit, in force since 29 April 2024 in the EU and 30 September 2024 in the UK, requires you to demonstrate timely affirmation. The sell-side legal teams won't risk a new vendor that hasn't been through their compliance process. MiFID II RTS 22 has a similar effect via the transaction reporting chain. It's not that regulators mandate MarkitWire (they don't), it's that the path of least resistance through compliance is the platform you already use.

The third is OSTTRA itself. CME and S&P together have a financial interest in MarkitWire's continuity. They have the relationships, the audit trails, the certifications, the integration libraries, the salesforce. A startup pitching the alternative doesn't.

Which is fine. It's the same shape as every infrastructure displacement story. Reuters had its turn, then Bloomberg, then Refinitiv. There was nothing inevitable about Bloomberg winning except a single decision in December 1981 about how much functionality to put on a single trader's terminal. The replacement of MarkitWire, if it happens, will turn on a single decision about what the unit of comparison is.

A note on what's coming

There's an obvious question hanging over all of this. If you reduced an OTC trade to its canonical economic terms, properly canonical, dealing with day-count aliases and calendar codes and decimal precision, and hashed those, two counterparties with the same trade would derive the same hash. Different terms, different hash. A comparison would tell you, in one operation, exactly which field differs.

That's not a new idea. ISDA's Common Domain Model, hosted at FINOS since 2022, takes a serious run at the canonical-form half of it. Several private startups have tried the hashing half. None of them, as of this writing, has put both halves together in a way the buy-side ops team can use without changing the OMS.

The interesting question is no longer whether the abstraction is right. The interesting question is who builds the version that runs on a Tuesday morning, in a real ops queue, against MarkitWire's installed base, without forcing the dealer to do anything different.

Anyway.