Tech Due Diligence Checklist for Trading Platforms
If you are buying or investing in a trading platform, the demo will look excellent. Order books render, prices tick, and the API answers quickly. The tech due diligence checklist for trading platforms below covers what that demo cannot show: matching correctness, order integrity, reconciliation, market-data handling and the regulatory evidence a data room must hold.
Key takeaways:
- Generic M&A technology checks cover code, security and team shape. They do not test matching or reconciliation.
- Ask for latency percentiles at order entry, matching and market-data fan-out, not an average from a demo test.
- Prove order integrity by replaying the event log and reconciling every fill, fee and balance movement.
- Treat market-data gaps and failover behaviour as valuation questions.
- FCA permissions, immutable audit trails and PSD2 evidence belong in the data room before close.
Why a Generic M&A Tech Checklist Fails on a Trading Platform
A generic M&A technology checklist asks whether the code is maintainable, security holds, and the team can carry the platform forward. Those checks can all pass on a system that produces incorrect fills under concurrency, leaves the ledger short at end of day, or drops market-data sequences without telling anyone.
That gap is why a technology due diligence checklist for M&A gets you most of the way on a SaaS business and only part of the way here. Trading software carries correctness properties ordinary web platforms do not. An order is a state machine with money attached. A fill is a financial event that must be reproducible years later. A price feed’s integrity affects every downstream calculation, from margin to client statements. When one breaks, the damage appears in client balances, regulatory returns, and eventually the price an acquirer will pay.
The lens for the rest of this article is narrow. We look at latency and throughput under real load, order integrity and matching-engine correctness, reconciliation and ledger completeness, market-data handling and failover, and the regulatory evidence a regulator or acquirer will demand. Those five areas are where deals get repriced, renegotiated or abandoned, and none can be confirmed from a walkthrough of the user interface or a warehouse of architecture diagrams.
Latency and Throughput Due Diligence: Testing Under Real Load
Latency and throughput due diligence means measuring percentile response times at order entry, at the matching engine and at market-data fan-out under peak and burst conditions, then checking that the platform degrades in a way the business understands. An average latency figure from a demo load test tells you little about capacity.
Ask for p50, p95 and p99 numbers for each stage of the path, not a single blended figure. The tail matters most: a platform with a comfortable median and a p99 that spikes whenever the book gets busy will lose market makers first. Ask what order mix the test used. Real trading is cancel-heavy and amend-heavy, and a generator that fires only new limit orders will miss the queue contention that hurts.
Separate peak from burst. Peak is a busy Tuesday. Burst is a news event, an open, or a liquidation cascade arriving in seconds. Ask what happened the last time the platform saw one, and what the on-call team saw. If you need a short sequence for gathering this evidence, our 14-day investor tech due diligence checklist sets out the order to request it in. A test against a single node on a warm cache is a starting point, not proof of production capacity.

Order Integrity and Matching-Engine Correctness
Order integrity checks confirm that every order follows a defined lifecycle, that partial fills, cancels and amendments are applied once and in the correct order, and that price-time priority, self-trade prevention and restart behaviour can be shown from evidence.
Start with the state machine. Ask for the diagram, then for tests that exercise every transition, including invalid ones. A cancel that arrives after a fill should be rejected cleanly, and an amendment should either preserve queue position or lose it according to a documented rule. Look for idempotency on cancel and amend calls; duplicate submissions from a flaky client connection are routine, and the engine must absorb them without creating phantom orders.
Concurrency is where most matching engines reveal themselves. Put orders at the same price from several sessions at once and check the sequence numbers. Restart the matching node and confirm the book rebuilds to the same state from the event log. Ask whether any incident of divergent books exists, and what was done. The engineering team should answer without searching chat. Our tech M&A due diligence CTO playbook covers how to structure this review across a short deal window. Undocumented matching rules are a common finding, and if a rule exists only inside one engineer’s head, the acquirer inherits a liability.

Trade Reconciliation and Ledger Completeness
Reconciliation due diligence proves that every trade, fee, commission and balance movement can be traced across the matching engine, the ledger and downstream reporting, and that the platform can replay a trading day from its event log and reach the same numbers.
Ask for end-of-day break reports for the last several months rather than a summary slide. Read the breaks. Ask what a typical break looks like, who clears it, how long it stays open, and whether the same break repeats. A few aged, unexplained breaks are a worse signal than many same-day items that get closed.
Check that replay is possible. Take one day, pull the event log, rebuild the state, and compare the output to the books. The log must be complete, ordered and retained long enough to cover the regulatory and audit horizon. Then test the edges: fee tiers, rebates, funding payments, currency conversion and any corporate actions. Every one is a place where a ledger quietly drifts.
A clean reconciliation report should show matched trade counts, matched value, unmatched items with their age, root cause of each, and a named owner. If reconciliation has only ever been run in a spreadsheet by one person, the acquirer is buying a process that does not survive the handover.

Market-Data Handling, Feeds, and Failover
Market-data checks cover how feeds are ingested, what happens when a sequence gap appears, how a snapshot restores state, how clocks are synchronised, and how the platform behaves when a primary feed or a matching node fails.
Ask for gap detection first. A feed handler that quietly continues after a missed sequence will price instruments from a stale book. The right behaviour is detection, alerting, and recovery from a snapshot plus incremental updates. Confirm that duplicates are suppressed and every message carries a source timestamp the platform can compare against its own clock.
Time synchronisation matters. Check which time source is used, how drift is monitored, and how timestamps feed into trade records and audit trails. On failover, ask what the standby node holds, how long promotion takes, and whether trading halts or continues while data is stale. Both answers can be defensible if the choice is deliberate and documented. A feed that silently freezes during a failover is a valuation issue, because client statements, risk limits and regulatory reporting all sit on top of it.
Regulatory Evidence the Data Room Must Contain (FCA, PSD2, Audit Trails)
The data room should hold immutable audit trails, evidence of FCA permissions and regulatory reporting, PSD2 strong customer authentication failure and contingency documentation, and a clear GDPR data map. When those artefacts are missing, the acquirer carries the remediation cost.
Audit trails need to be tamper-evident, queryable and retained for the period the business is held to. Ask to see a sample export that reconstructs a specific client action from end to end, including who approved it and on what system. For an FCA-regulated acquirer, expect permissions, reporting returns, complaints records and incident history to be indexed and current.
On payments, the PSD2 implementation CTO checklist sets out the SCA failure modes and contingency artefacts UK fintechs are asked for. In diligence terms, you want the exemption logic, the fallback when authentication fails, and evidence the fallback has been tested rather than written down. Then GDPR: lawful basis for each data category, retention rules, processor agreements and the process for subject requests. Missing regulatory evidence rarely stops a deal outright. It moves value from the seller to the buyer.

Failure Patterns That Kill Deals After a Clean Demo
The recurring post-demo failures are untested peak load, ledgers that do not reconcile, matching rules that exist only in code, single points of failure in the trading path, and missing regulatory evidence. Each one is visible during diligence if you ask for the right artefact early enough.
Untested peak load appears the first time a market event arrives, usually within weeks of completion. An unreconciled ledger appears when finance chases a break nobody owns. A matching rule that lives in one engineer’s memory disappears when that engineer leaves, which is common after an acquisition. A single point of failure in the order path shows up at three in the morning, when the person who understood it is asleep.
Regulatory gaps surface last and cost the most, because they arrive as a letter rather than a bug report. Our ZuluTrade CTO crisis turnaround case study describes a trading platform where daily crashes and siloed engineering sat behind a surface that looked functional to anyone outside the team. Spotting these patterns is a diligence activity. Fixing them after close is a valuation event.
What This Means for the Deal
A trading-platform technical due diligence checklist replaces the generic M&A list. The checks that decide the outcome are latency and throughput under real load, order integrity and matching correctness, reconciliation and ledger completeness, market-data handling, and the regulatory evidence held in the data room. None can be demonstrated by a clean demo. Investors and acquiring CTOs should treat a missing reconciliation trail, unproven peak capacity or absent FCA and PSD2 artefacts as valuation issues first, engineering issues second.
FAQ
What is the difference between a generic tech due diligence checklist and a trading-platform one?
A generic checklist tests code quality, security and team structure across any software business. A trading-platform checklist adds matching-engine correctness, order lifecycle integrity, reconciliation and ledger completeness, market-data handling, and the regulatory evidence a regulated acquirer needs. Those tests apply where an order carries money and a fill must be reproducible.
How do you test latency and throughput during due diligence without disrupting the live platform?
Run the test against a production-like environment with the same topology, or use shadow traffic that mirrors live order flow without affecting real books. Ask for p50, p95 and p99 figures at each stage, plus a separate burst test. Never accept results from a single node on a warm cache.
What regulatory evidence should a trading platform’s data room contain for an FCA-regulated acquirer?
Immutable, queryable audit trails with defined retention, current FCA permissions and reporting returns, complaints and incident records, PSD2 strong customer authentication failure and contingency documentation, and a GDPR data map covering lawful basis, retention and processor agreements. Each item should be indexed and traceable to a named owner.
How do you verify order integrity and reconciliation if the platform’s logs are incomplete?
Treat incomplete logs as a finding in their own right, since replay and audit both depend on them. Where gaps exist, reconstruct what you can from downstream systems, broker statements and ledger entries, then document the missing period. Ask the seller to price the remediation, or negotiate a holdback against it.
Written by Agents24x7