The Empty Chair Test: Can the File Explain a Compliance Decision After Its Owner Leaves?
Compliance decisions rarely fail in the moment they are made. They fail later, when the person who made them has moved on and the file no longer explains what happened, why it happened or who approved it. For UK firms operating under regulatory expectations of demonstrable governance, that is not an administrative inconvenience. It is an evidential weakness.
The empty chair test is a simple drill. Select several material compliance decisions taken at least 90 days ago by someone who is no longer readily available: they may have left, moved to another function, gone on extended leave or been made redundant in a restructuring. Then ask a reviewer who was not involved to reconstruct each decision from records alone. No phone calls. No corridor recollection. No reliance on the decision-maker's inbox. The question is whether the organisation can prove how and why the decision was reached, not whether someone can remember it.
Why the empty chair matters
Staff turnover is normal. Restructuring is normal. Extended leave is normal. What is not acceptable is a position where the firm's ability to explain a regulatory judgement depends on a single individual's continued availability.
In the UK, this pressure appears in several familiar places. Senior Managers and Certification Regime responsibilities do not evaporate when a manager leaves. Consumer Duty outcomes need to be evidenced, not merely asserted. Anti-money laundering controls must be demonstrable to supervisors. Data protection decisions may need to be explained to the Information Commissioner's Office. Companies House filings and board approvals may later be examined by lenders, auditors or counterparties. In each case, the record has to do the work that memory cannot.
The test is therefore not a memory quiz. It is an evidential resilience check. It exposes where decision logs are incomplete, where version control is informal, where delegated authority is assumed rather than documented, and where meeting notes record attendance but not challenge.
Designing the drill
A workable exercise does not need to cover every decision. It needs a defensible sample. Select five to eight decisions that are material, cross-functional or dependent on delegated authority. Good candidates include a policy exception, a customer remediation approach, a supplier risk acceptance, a model or rule change, a complaints handling decision with conduct implications, or a regulatory notification decision.
For each one, assemble the evidence pack without contacting the original decision-maker. Timebox the reconstruction to a reasonable period, such as two hours per decision. The reviewer should record what was found, what was missing and what required inference. Only after the reconstruction is complete should the firm compare the result with the decision-maker's own account, if one is available. That comparison is useful, but it is not the test. The test is whether the file stood alone.
Five evidence tests, plus retrievability
A practical review can use six criteria. Each is scored from 0 to 2: 0 where the evidence is absent or unusable, 1 where it is partial or requires significant inference, and 2 where it is clear, contemporaneous and independently intelligible.
Decision record. Does the file identify the decision, date, owner, trigger, options considered, rationale, approvals, conditions and review date? A decision log that records only the outcome is not a decision log. It is a minute of an event.
Version control. Can the reviewer identify which policy, procedure, model, template or customer file was in force at the time? Is there a version number, approval date and change history? Can the firm distinguish a draft from an approved document?
Delegated authority. Was the decision taken within a documented delegation? Is the authority matrix current, and does it address sub-delegation, conflicts and quorum? An email saying 'approved' is not the same as evidence that the approver held the authority to approve.
Meeting evidence. Do minutes or notes capture the substance of the discussion, including alternatives considered, challenge raised and dissent recorded? Attendance lists prove presence. They do not prove deliberation.
Rationale evidence. Is there a clear link between the decision and the regulatory obligation, customer outcome, risk assessment or data relied upon? Where advice was taken, is the advice identifiable and was it addressed? Where a judgement was made, is the basis for that judgement visible?
Retrievability. Could a reviewer find the pack without insider knowledge? Is it held in a controlled repository with consistent naming, access permissions and retention rules? Or is it scattered across shared drives, email folders and personal notes?
A simple scoring model
| Criterion | Score 0 | Score 1 | Score 2 |
|---|---|---|---|
| Decision record | No record or outcome only | Partial record | Complete and contemporaneous |
| Version control | Cannot identify version | Version identified but approval unclear | Version, approval and change history clear |
| Delegated authority | No evidence of authority | Authority implied | Documented and current |
| Meeting evidence | No notes | Attendance only | Challenge, options and dissent recorded |
| Rationale evidence | No rationale | Rationale asserted | Rationale linked to evidence and obligations |
| Retrievability | Scattered or inaccessible | Findable with help | Controlled repository and naming |
A total of 0 to 4 indicates critical fragility. The firm is likely relying on people rather than records. A score of 5 to 8 indicates a developing position: some evidence exists, but reconstruction relies on inference. A score of 9 to 10 is adequate, provided gaps are remediated. A score of 11 to 12 is resilient. The purpose is not to produce a league table of individuals. It is to identify where the organisation's evidence chain breaks.
From score to action
Remediation should be practical. Simplify decision templates so they capture rationale rather than just approval. Add mandatory fields to decision logs. Move material records into a controlled repository with retention rules. Require handover packs for departing staff that include live decisions and open conditions. Review delegation matrices at least annually and after restructuring. Test a rolling sample each quarter, not once a year.
The board or compliance committee should see the aggregate result. A pattern of weak retrievability or missing rationale is a governance signal. It suggests the firm is comfortable with informal assurance and uncomfortable with documented proof.
The behavioural trap
Most firms do not set out to rely on memory. They simply allow convenience to fill the gap. A decision is made in a meeting, confirmed by email, stored in a folder and understood by the people in the room. That works until the room changes.
The empty chair test is deliberately uncomfortable because it asks a simple question: if the decision-maker vanished tomorrow, would the file still speak? For UK compliance teams in 2026, that is a more useful measure of maturity than another policy refresh. Defensible compliance depends less on who remembers a decision than on whether the firm can prove how and why it was made.
