98.8% Is a Market Signal. Your Perpetual KYC Pilot Still Needs an Exit Test.

On 2 September 2026, Liminal reported that 98.8% of banks in its KYC in Banking research had perpetual KYC live, in rollout, or planned within the following twelve months. That combined measure is a strong signal of buying intent. It does not mean 98.8% have completed a successful implementation.

For management companies, fund administrators and trust and company service providers, the useful response is to define what a successful pilot must prove. A banking survey can inform a purchasing conversation. It cannot establish whether a particular firm’s customer structures, jurisdictions and review responsibilities are covered.

Start with the denominator

Consider a hypothetical pilot covering 200 customer relationships. All 200 appear in a dashboard, but only their directly registered entities are connected to a registry feed. Intermediate holding companies and individual beneficial owners remain outside the monitoring population. Reporting 100% customer coverage would obscure the limitation that matters most.

A better pilot records the entities and people that require monitoring, the sources available for each, and the specific changes those sources can reveal. A registry that reports directors may provide little useful information about control exercised through a trust. Coverage needs to describe the evidence available, not merely count enrolled customers.

The acceptance criteria should therefore distinguish relationship coverage from attribute coverage. Ownership, directorships, authorisation status and identity information may each have different sources and update frequencies. An unavailable source should remain a visible exception with an assigned alternative control.

Test the events that should have been found

A live demonstration usually starts with a change that the software can detect. A stronger assessment starts with independently known changes and asks whether the system found them. Use an authorised test environment and a small, documented set of historical or simulated events relevant to the firm’s business.

Include an ownership transfer, a corrected name, an irrelevant address-format change and two customers with similar names. Record whether the right relationship was identified, whether the event was classified sensibly and whether the resulting review reached its owner. These are proposed acceptance tests, not claims about the capabilities of any particular vendor.

A missed event deserves investigation even when the overall alert count looks healthy. The cause might be a delayed registry filing, an incorrect entity identifier, a broken integration or a rule that filtered out a meaningful change. Each requires a different remedy. Buying a broader feed does not automatically fix a matching problem.

Deliberately interrupt a feed

A monitoring service can appear quiet because nothing changed or because observation stopped. A pilot should demonstrate that the firm can distinguish those states. In a controlled test, interrupt a source and inspect how the gap appears to the team.

The record should show the last successful check, the affected population and the person responsible for the fallback. On restoration, test whether missed events are recovered and whether duplicated alerts are handled. Recovery should leave a trace that explains the period of reduced visibility.

This is an operational design recommendation. It does not imply that every customer needs every possible source or that all checks must run in real time. The appropriate frequency depends on the relevant risk, the service and the information that can actually be obtained.

Measure the whole review

A useful pilot scorecard follows work through to a supported decision. Detection speed matters, but so do time awaiting assignment, time awaiting evidence and the age of unresolved material cases. Measure analyst effort per completed review rather than per alert dismissed.

Review a sample of closures with a colleague who did not handle the original case. Can that person identify the change, the evidence considered and the reason for the outcome? If they need an oral explanation, the workflow still depends on knowledge outside the record.

Before expansion, agree how failures affect the rollout decision. A missing ownership source in a material jurisdiction may justify keeping that population on an alternative process. A minor usability issue may be tolerable with training. Treating both as generic backlog items hides the actual decision.

Make expansion conditional on evidence

Fidify’s earlier discussion of perpetual KYC readiness focused on structured customer data. The next step is to demonstrate that the data supports a complete monitoring process under normal conditions and when a component fails.

Liminal’s update gives firms a reason to examine their plans now. The pilot should give them their own reason to proceed: a defined population, tested triggers, workable fallbacks and decisions another reviewer can reconstruct. Scale follows that evidence.

Sources and further reading

Industry update: Liminal, Perpetual KYC has moved from if to how, 2 September 2026. Survey findings are attributed to Liminal; this article does not treat the combined live, rollout and planned figure as completed adoption.

Related Fidify analysis: Perpetual KYC: What It Means and Whether Your Firm Is Ready.