Purchasing Guides / Trade Confirmation Matching

Trade Confirmation Matching

A purchasing guide to buying a workflow that reads counterparty trade confirmations, compares their terms with your booked trades and lists matches and breaks for your operations team.

Yearly running cost

Assuming 6,000 confirmations a year

By hand: $36K / year; 800 hours of work
Subscription: $2.4K+ / year
This workflow: $281 / year; Machine usage only; setup, hosting and review are extra.

Where should the confirmations come from?

Compare the economic terms on each confirmation with the trade you booked. Review the breaks instead of reading each confirmation line by line.

Maximum document size: 20 pages per confirmation.

Match uploaded confirmationsConfirmations and a trade export, with a break queue$1,224 – 3,240 setupRead from my mailbox and trade systemRepeat matching from approved sources$2,088 – 5,640 setup

Your operations team decides every break. No trades placed, affirmed, amended or cancelled, and no payments or settlement instructions.

What is included

Read agreed confirmation formats, extract economic terms, find the booked trade and return matched, break and check-by-hand results with page references.

Start with your confirmation formats, a trade export with stable trade identifiers, agreed tolerances for price, rate, amount and dates, and operations-checked examples. The connected option reads a confirmations mailbox and your trade records; it does not affirm, amend or cancel trades.

How reliable should the matching be?

A high match rate is not enough. Choose how accurately terms must be read, how many real breaks it must catch and how few false breaks it may raise.

StandardAccurate terms, with uncertain confirmations held for review.Terms read correctly: ≥98%Real breaks caught: ≥97%IncludedStrictMore tests on difficult counterparties, amendments and term types.Terms read correctly: ≥99.5%Real breaks caught: ≥99%Setup +$360 – 960My own targetsAgree your own acceptance requirements with the provider.Quote separately

How these standards are measured

Test on held-out confirmations from counterparties and periods not used for tuning. Measure term extraction, breaks found and false breaks separately. Amended, duplicate, unreadable and wrong-counterparty confirmations must stay visible; missing output is a failure.

Targets for your selected standard
What is checked	Target
Terms read correctlyA term passes only when value, currency, unit, date and party are all correct; an unreadable term counted as unreadable is correct.	≥98%
Real breaks caughtBreaks flagged or held for review divided by all operations-labelled breaks; missing results count as misses.	≥97%
Breaks that are realFlagged breaks that operations confirms divided by all flagged breaks.	≥90%
Complete source linksResults must keep the confirmation document, page, booked trade identifier and tolerance rule version.	≥100%

Compare prices, rates, amounts and dates with exact arithmetic and agreed tolerances. A model reading is not evidence that a trade agrees; the result must cite the confirmation page and the booked trade.

How quickly should a confirmation be checked?

Choose how quickly each confirmation should be ready for review. Your team's investigation and counterparty follow-up are separate.

Within 2 minutesIncludedWithin 60 secondsSetup +$144 – 480Within 30 secondsSetup +$288 – 840

Timing details

Time from receiving a complete confirmation to a stored match result, including reading the document, finding the booked trade, comparison, queueing and retries. Confirm document size and hardware with your provider.

The target applies to at least 95% of agreed test runs, with 4 in progress at a time.

How much do you want to spend per confirmation?

Choose the machine budget for reading a confirmation and comparing its terms. Tighter budgets may use smaller models or fewer checks.

Up to $0.25IncludedUp to $0.15Setup +$72 – 360Up to $0.10Setup +$216 – 600

Cost details

Includes model calls, retries and shared hosting allocation. Human review, trading and settlement systems, matching platforms and the calling agent are separate. Failed and repeated runs also consume resources.

Reference machine cost: $0.09 – 0.17 per confirmation at 500 confirmations a month. The selected cap is a target to test, not a replacement for this estimate.

Reference workload: 4 pages per confirmation.

Where do you want it to run?

Keep confirmations and trade records in your cloud or on your own server. Choose whether confirmation pages may go to an approved AI service.

Run it onAI model access

Data and access details

Runs in a cloud account you control, with access controls and logs.

Only approved confirmation pages and trade fields go to the selected external model. Agree access and retention first.

Use read-only access to trade records and mask account details where possible. No external calls uses uploaded files and private inference; it cannot use the connected mailbox and trade-record option.

How do you want to use it?

Use a review page, your current AI agent or a dedicated matching agent. You can choose more than one.

A review web pageUpload confirmations and a trade export, review matches and breaks, and export the results.IncludedMy existing AI agentAsk your agent to submit confirmations and retrieve match results and open breaks.Setup +$72 – 240A dedicated agentUse a dedicated agent to run matching and explain why a confirmation did not match.Setup +$144 – 480

Anything else your provider should know?

Optional. Your choices are included automatically.

Confirmations per monthYour setup budget (USD)

Your confirmation formats, asset classes or trade system

Common questions

Do I need this if we already use an electronic matching platform?

Not necessarily. Keep the platform for trades it already matches electronically. Buy a workflow when paper, PDF or email confirmations still need manual checking.

Does a match mean the trade is affirmed?

No. It is a comparison backed by the confirmation and your booking. Your team affirms, disputes or corrects trades in your existing process.

Which trade types can it handle?

The ones whose confirmation formats and terms you agree with the provider. Additional asset classes, languages or structured products can change the scope and price.

Can it contact counterparties about breaks?

No. It can draft an internal break note. Your team decides whether and how to contact the counterparty.

Can I run it without an external AI service?

Yes, with uploaded files. Choose private processing and supply your own hardware. Measure model quality and running cost during the pilot.