VAVaultgridplus
Guides › Compliance

Eight Trade Surveillance Buying Mistakes That Let Market Abuse Slip Through

Most surveillance failures start in procurement, not operations. Eight buying errors that leave spoofing, cross-venue manipulation and insider dealing undetected, and what to write into the contract instead.

Vaultgridplus editorial team · November 2026 · 8 min read

Why Surveillance Gaps Usually Start in the RFP

When a regulator finds that a broker missed a layering pattern or an insider trade, the root cause is rarely a lazy analyst. It's usually a decision made a year or more earlier: the firm bought trade surveillance software that couldn't see the right data, couldn't be tuned to its actual flow, or kept trade and communications alerts in separate systems that nobody joined up.

The legal standard is broadly similar across major markets. Under EU MAR Article 16 and its UK onshored version, firms that professionally arrange or execute transactions must have effective arrangements, systems and procedures to detect and report suspicious orders and transactions. The supporting EU technical standard (Delegated Regulation 2016/957) covers orders, including cancellations and modifications, not just fills. India's SEBI expects stock brokers to act on exchange-generated alerts and to run their own surveillance. Regulators such as Hong Kong's SFC, Singapore's MAS, Australia's ASIC and US bodies applying FINRA supervision rules all expect controls proportionate to the business and demonstrably effective.

Advertisement

"Effective" and "proportionate" are where procurement gets tested. Below are eight mistakes that weaken compliance most often, each with what to require from vendors.

Accepting Out-of-the-Box Thresholds Never Calibrated on Your Data

Vendor demos run on clean sample data with thresholds tuned to a generic equities desk. Your desk isn't generic. A small-cap market maker, a fixed income agency broker and a quant asset manager produce very different order-to-trade ratios and cancellation rates. Default parameters will either bury analysts in false positives or, worse, sit silent while real patterns pass just under the line.

Supervisors tend to press firms on why a threshold sits where it does. A number copied from a vendor manual isn't a defensible answer.

Require instead:

  • A proof of concept on at least three to six months of your own historical orders and trades, not vendor sample data.
  • Documented calibration methodology for each model, including above-the-line and below-the-line testing to show what the thresholds miss.
  • A tuning process you can run yourself, with versioned parameter changes and an audit trail.
  • A committed number of calibration cycles in year one, priced into the statement of work.

Feeding the System Executions Only, Without Orders and Quotes

This is the most common and most damaging gap. Spoofing and layering detection depends on seeing orders that were placed and cancelled before they executed. If surveillance receives only executed trades from the back office, those models have nothing to analyze. Quote stuffing, momentum ignition and marking the close leave their main footprint in order and quote activity too.

Timestamp precision matters as well. Layering can play out in milliseconds. If your order management system rounds to the second, or the vendor normalizes timestamps across venues badly, the sequence of events gets scrambled.

Require instead:

  • Ingestion of the full order lifecycle (new, amend, cancel, reject, fill) at millisecond or finer granularity.
  • Market data, top-of-book at a minimum and depth where you trade thin markets, so alerts can be judged against the market at the time.
  • A data completeness dashboard that reconciles record counts against source systems daily and flags dropped feeds.
  • Written confirmation of which models degrade, and how, when order data is missing.

Buying Single-Venue, Single-Asset Coverage

Manipulators work across venues and instruments because that's where single-venue surveillance goes blind. The classic patterns: pushing a cash equity on a thin venue to profit from a derivative position, spoofing on one lit venue while executing on another, or trading a bond to move a CDS price. An exchange sees only its own book. Your firm may be the only party that sees both legs.

EU and UK MAR expect surveillance to reach related instruments, and Indian regulators have pushed brokers toward monitoring across cash and derivatives segments. A system licensed per asset class, with equities live and derivatives "phase two," leaves a window that can stay open for years.

Require instead: a single data model linking instruments to their underlyings and related derivatives, cross-venue order book reconstruction, and cross-asset scenarios in the initial scope rather than the roadmap. Ask the vendor to show a cross-product alert on your PoC data, not a slide.

Keeping Trade and Communications Alerts in Separate Systems

Many firms run trade surveillance from one vendor and e-comms or voice surveillance from another, with separate analyst teams and case queues. The result is predictable. A trade alert on an unusual pre-announcement position gets closed as "consistent with strategy" while, in the other system, a chat message from that trader about the deal sits unreviewed.

Insider dealing and collusion cases usually need both sources to establish intent, and supervisors expect firms to be able to reconstruct a trader's activity as a whole.

Require instead: either one platform covering both (vendors including NICE Actimize, Behavox and Global Relay market combined offerings, with differing depth), or a documented integration that links alerts by trader ID, instrument and time window into one case. Test it: hand the vendor a trade alert and ask how many steps it takes to see that trader's communications in the 48 hours before. If the answer involves switching systems, treat that as a warning sign.

Signing Up for Black-Box Models You Can't Explain to a Regulator

Machine learning and behavioral analytics can cut false positives. They also create a new problem: if your compliance officer can't explain why an alert fired, or why one didn't, the firm can't defend its control framework. Some regimes apply model risk expectations to surveillance analytics, and many banks put them under internal model governance anyway.

Red flags: the vendor can't show the features driving a score; only the vendor's data scientists can tune the model; there's no way to run a rules-based scenario alongside the model and compare results.

Require instead: plain-language explanations attached to each alert, model documentation your validation team can use, the ability to run challenger rules in parallel, and contractual access to model change notes before any update reaches production.

Treating Best Execution Monitoring as an Afterthought

Best execution monitoring and market abuse surveillance share much of their data but answer different questions. One asks whether clients got the best possible result; the other asks whether someone manipulated the market. Firms often assume the abuse platform "does best ex," then find it offers only basic slippage reports, with no venue analysis, no fixed income benchmarks and little that evidences compliance under MiFID II, UK FCA rules or equivalent obligations elsewhere.

The reverse happens too. A TCA tool bought for best execution gets expected to catch front-running. It usually can't.

Require instead: a scope statement listing which best execution outputs are included (benchmark types, asset classes, venue comparison, exception workflows) and which need a separate tool. If both functions come from one vendor, confirm they share the same normalized data so you don't pay twice for ingestion.

Underestimating Case Management and Suspicious Order Reporting

Detection is half the job. Under EU and UK MAR, suspicious orders and transactions must be reported to the competent authority without delay through a STOR. Indian brokers must report suspicious activity to exchanges within set timelines, and the SFC, MAS and ASIC each have their own reporting expectations. If analysts track alert decisions in spreadsheets and draft STORs in email, your audit trail is weak however good the models are.

Require instead:

  • Case management with mandatory rationale fields, four-eyes escalation and immutable audit logs.
  • STOR or local-equivalent templates for each jurisdiction you report in.
  • Management information on alert volumes, aging, closure reasons and escalation rates, which is what a regulator will ask for in a review.
  • Record retention configurable to the longest regime you operate under; five years is a common minimum, and some firms keep seven or more.

Negotiating Price Without Pinning Down the Cost Drivers

Surveillance pricing usually scales with message volume, asset classes, venues and seats. Firms that sign on a headline annual fee often find that adding a derivatives venue or a jump in order volume triggers a renegotiation. The table shows indicative typical 2026 market ranges. They vary widely by contract and region, so confirm with vendors.

Firm profileTypical annual licence (USD)Implementation (one-off)Main cost driver
Small broker or asset manager, one or two asset classes$60k–$200k$30k–$120kAsset classes and seats
Mid-sized multi-venue broker$200k–$700k$150k–$500kOrder message volume, venues
Tier-1 bank, multi-asset, multi-region$1M–$5M+$1M+Data volume, jurisdictions, comms add-ons
E-comms and voice add-on$50k–$400k$40k–$250kMonitored users, languages

Before you sign, get volume bands with fixed overage rates, a price for each additional venue and asset class, regulatory rule updates included in the base fee, and exit terms that return your normalized data and case history in a usable format.

Your next step, before issuing any RFP: map every order, quote, trade and communications source against each regime you report under. Any vendor that can't ingest that full map during a PoC on your own data should leave the shortlist, whatever the demo looked like.

Questions buyers ask

Do EU and UK MAR require surveillance of orders, or only executed trades?

Orders as well. The EU technical standard supporting MAR Article 16 explicitly covers orders, including cancellations and modifications, and the UK kept the same approach after Brexit. A system that monitors only executions will struggle to show it can detect spoofing and layering, which undercuts any claim that your arrangements are effective.

How long does a trade surveillance implementation usually take?

A single-asset, cloud-hosted deployment for a smaller firm often takes a few months. Multi-asset, multi-region bank programs can run a year or two, mostly because of data sourcing and normalization rather than software configuration. Timelines vary by vendor and data quality, so build a parallel-run period into the plan.

Is a combined trade and communications surveillance platform always better?

Not always, but you need the alerts linked one way or another. A single platform simplifies case management and correlation; best-of-breed tools may be stronger in particular asset classes or languages. If you choose separate vendors, make documented integration by trader, instrument and time window a contractual deliverable.

Can we rely on exchange-generated alerts instead of our own surveillance?

Generally no. Exchanges see only their own venue, not your cross-venue or cross-asset activity or your client relationships. SEBI, for example, expects brokers to act on exchange alerts and also run their own surveillance, and many other regulators take a similar line. Confirm the specifics for your jurisdiction with compliance counsel.

What should a proof of concept for trade surveillance software include?

Use at least three to six months of your own order, trade and market data, including known historical cases if you have them. Measure alert volumes, false-positive rates, below-the-line misses and the time analysts take to work an alert. Also test data ingestion failures, because how the system behaves when a feed drops says a lot about operational risk.

Written by the Vaultgridplus editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.