Checking afterwards is not checking.

The records were complete. The product did not match what the customer had ordered. When a sign-off happens hours after the work, you have accounting, not verification.

A print facility had every quality system you would expect. Standard procedures, job records, in-process checks, sign-off sheets, customer specifications. All present. What was missing was the discipline of using them as real controls rather than as paperwork completed alongside the work.

Job records were filled in at the end of the shift, not during the run. Defect reports were raised after the job had moved on. Colour checks, and checks that the printed layers lined up, were scheduled at set intervals and the record showed the correct times, but in practice they clustered at convenient moments and were signed off afterwards to match the schedule. The checking apparatus was there. The checking was not.

Nobody introduced the shortcuts on purpose. The first one was taken once, under real time pressure, on a job that was plainly fine. Nothing went wrong. So it was taken again. Diane Vaughan, who spent years inside the NASA record after Challenger, called this the normalisation of deviance: small acceptances, each one invisible on its own and fatal together. It settles into a production line the same way it settled into a launch programme.

The cost is measured. The Get It Right Initiative puts the directly measurable cost of avoidable error in UK construction at around 5% of project value, and its best estimate with hidden costs included at around a fifth. The reading I take from their work is about timing. When the systems that catch error are built to find it afterwards, it is already too late.

The inspector signing hours after the work is not dishonest. She is working in a system that treats a record written later as equal to a record written at the time, and that measures whether the record exists rather than whether it is true. Measure completion and people optimise for completion.

The fix was not training people to care more. It was redesign. The check became the gate between one stage and the next, not a parallel task to do when convenient. The job record was completed at the moment of the action it described. And the daily review stopped asking whether the records were complete and started asking whether they were accurate, with section leaders walking recent records against the product coming off the line.

The share of product passing every stage first time rose from 68% to 76% across the two quarters either side of the change. About three points of that came from a shift in product mix. The other five came from the redesign, and they show product by product.

In your operation, which checks are signed at the time, and which are signed later to match the schedule?