Jenesys Associates All articles
Operational Improvement

The Second Signature: Proving Your Review Controls Still Catch Anything

Jenesys Associates
The Second Signature: Proving Your Review Controls Still Catch Anything

Photo: Lars Jacob for CabarEng., Public domain, via Wikimedia Commons

The second signature on a payment run. The manager's approval on a contract variation. The peer review before a report leaves the building. Four-eyes checks are among the most common controls in UK operations and among the least examined. They are cheap to draw on a process map and expensive to run properly, which is precisely why so many of them drift quietly into ceremony.

The design intent is simple: two independent judgements, the second of which is informed enough to disagree. The operating reality is often a screen that opens with the decision already taken, a queue that must be cleared before midday, and a checker who has reviewed the same colleague's work every week for three years. Under that combination, the second pair of eyes stops seeing and starts confirming.

When a check becomes a click

Several forces push a review control towards ritual. Volume is the most obvious: when a checker is expected to clear ninety items before Friday, the rational behaviour is to approve anything that looks familiar. Familiarity compounds it. So does unclear criteria, where nobody can say precisely what the checker is checking against. Deference to seniority plays a part, as does the absence of consequence, where an exception raised leads nowhere and an exception missed is never traced back.

None of these are failures of individual diligence. They are design failures, and unlike individual lapses they scale. A weak check applied to thousands of transactions does not degrade gracefully; it converts the appearance of assurance into a false sense of it.

What a working control leaves behind

A functioning review control produces artefacts, and those artefacts can be inspected. There are four to look for.

First, evidence. A real check generates a record of what was examined and why the reviewer was satisfied, or a note of the exception that was raised. Second, criteria. The reviewer can state the standard against which the item was judged, in a sentence, without reaching for a procedure document. Third, escalation. Exceptions have a named destination and a decision is recorded there. Fourth, correction. Repeated exceptions change something upstream, whether that is a form, a training note or a supplier conversation.

If an organisation cannot point to all four, the control is not operating as designed. It may still be catching the occasional error by luck, but it is not a control in any meaningful sense.

Four tests worth running

The exception rate

Take twelve months of data and ask how many exceptions the control has generated. A review that has never disagreed with a first review is either operating in a flawless process or not operating at all. The second explanation is far more common. Cross-check against errors found downstream: credit notes, payroll corrections, complaints, reconciliation adjustments. If errors surface after the control, the control is not working.

The sample and the interrogation

Pull thirty items from the last quarter and ask the checker to explain, for three of them, what they compared and what would have caused them to stop. Ask when they last raised a concern and what happened next. Vagueness here is diagnostic. A reviewer who cannot describe their own last disagreement is not reviewing.

Re-performance or a controlled error

Ask a third party to independently re-perform twenty items and compare conclusions. Where governance permits, plant two or three realistic errors in a live batch with the knowledge of the process owner and the audit function, then observe whether they are caught. This is uncomfortable to arrange and considerably less uncomfortable than discovering the failure during a regulatory visit.

Follow the escalation

Trace the last five exceptions from detection to resolution. How long did each take? Who decided? Did anything change as a result? An escalation route that has never been used is a diagram, not a mechanism.

Triaging the controls that matter

No operation can strengthen every check. The practical move is to rank controls by consequence and by volume, then act on the top of the list. Consequence means financial exposure, regulatory obligation, safety or reputational damage. Volume means how often the control is relied upon.

Three responses follow. Retire the controls that sit low on consequence and high on effort, and say plainly that they are being retired. Automate where the test is objective: matching, tolerance checks and validation rules do this work more reliably than a tired reviewer at month end. Concentrate human attention on the small number of judgements that genuinely need a second opinion, and give those reviewers the time, criteria and authority to say no.

A thirty-day reset

Week one: inventory the review controls, name the owner of each and rank by consequence. Week two: sample-test the top ten against the four artefacts above. Week three: rewrite the criteria and escalation route for the three that matter most, and agree what evidence the reviewer must leave. Week four: report to the relevant governance forum with a rework measure attached and a re-test date in the diary.

The measure to add to monthly reporting is rework generated by the control: exceptions raised, corrections made, upstream fixes completed. It is a modest metric, but it separates a control that is doing work from one that is merely present.

The purpose of four eyes is not to distribute blame. It is to catch what one pair missed. That is a testable claim, and it should be tested on a schedule rather than assumed at design stage and never revisited.

All Articles

Related Articles

The Post Has Moved: A Preventative Audit for Role Drift

The Post Has Moved: A Preventative Audit for Role Drift

Permission Without Purpose: Retiring the Inherited Sign-Off

Permission Without Purpose: Retiring the Inherited Sign-Off

Learning by Osmosis: Auditing the Onboarding a New Starter Really Receives

Learning by Osmosis: Auditing the Onboarding a New Starter Really Receives