Standard Work Adherence Audit: How to Tell If Your Improvement Is Quietly Slipping

The Standard Work Adherence Audit does one job. It verifies that operators follow the defined standard work: the sequence, the time, and the in-process stock. You observe the work as it happens, record what you see against the standard, and let the tool calculate how far apart the two have drifted.

It sits in the Sustain phase of RDMAICS for a reason. By the time you reach Sustain, the improvement has been designed, tested and handed over. The control plan is signed. The dashboard is green. None of that tells you whether the people doing the work are still doing it the agreed way. Standard work decays quietly. A step gets skipped because it felt redundant. An approval drifts because the approver is busy. Nobody reports it, because nobody thinks of it as a problem.

Reach for this tool on a schedule, not when something breaks. A monthly or quarterly audit on a handed-over process is the point. You are not hunting for a fault. You are measuring the gap between the documented standard and the lived practice, so you can act while the gap is still small.

Everything that follows is built on the sample data that loads with the tool. The case is fictional. It describes an invoice processing flow with five steps, audited after the improvement was handed to the operating team.

The five steps are receive invoice, OCR capture, three way match, approve, and post payment. Each has a standard time agreed during the Improve phase. The auditor sat with the process, timed each step, noted whether the sequence was followed, and recorded an adherence percentage with a short comment.

The reason this example is useful is that nothing dramatic has happened. No step was skipped. No sequence was broken. If you only looked at output, you would see a process working. The audit still finds something worth acting on.

The input table asks for one row per step of the standard work. For each row you supply the step name, the standard time, the observed time, whether the sequence was followed, an adherence percentage, and a comment.

In the sample, receive invoice carries a standard time of 2 minutes and an observed time of 2.1 minutes. OCR capture is 3 minutes standard against 3.5 minutes observed. Three way match is 1 minute against 1.0 minute. Approve is 5 minutes against 6.2 minutes. Post payment is 2 minutes against 2.0 minutes. Sequence followed is marked Yes on all five.

The adherence column is where judgement enters. Receive invoice, three way match and post payment are recorded at 100 per cent with the comment Conforming. OCR capture is 86 per cent with the comment Slight delay. Approve is 81 per cent with the comment Approver slow. The comment field matters more than it looks. A percentage tells you a step has slipped. The comment tells the next reader why, months later, when memory has faded.

The tool returns five headline figures. Steps observed is 5. Average adherence is 93.4 per cent. Sequence followed is 5 of 5. Standard time total is 13 minutes. Observed time total is 14.8 minutes.

Read those together, not separately. Sequence followed at 5 of 5 is genuinely good news. The shape of the work is intact. Nobody has invented a workaround or reordered the steps. That is the failure mode that usually destroys an improvement, and it has not happened here.

Average adherence of 93.4 per cent is the figure that invites a shrug. It sounds close enough. The step level view is what stops you shrugging. The tool lists the two steps below 100 per cent: approve at 81 per cent with the comment Approver slow, and OCR capture at 86 per cent with the comment Slight delay. The average is being held up by three conforming steps. The loss is concentrated, not spread.

The time totals make the same point in a different unit. A 13 minute standard is being delivered in 14.8 minutes observed. That difference is not shared evenly. Approve alone accounts for 6.2 minutes against a 5 minute standard. The tool also offers an AI review that explains the calculated result in plain language if you want a written summary to paste into a report.

Drift has a signature. This audit is designed to show it before the output metrics do.

The first signal is a widening gap between standard time total and observed time total. In the sample that gap is 13 minutes against 14.8 minutes. On its own, in one audit, that is a data point. Across three audits with the gap growing each time, it is a trend, and the standard is no longer the standard.

The second signal is adherence concentrating in one step. Approve at 81 per cent is a single named step repeatedly underperforming. That is more actionable than a general decline, and it is more likely to spread, because downstream steps start absorbing the delay and inventing buffers.

The third signal is the one this sample does not show. Sequence followed is 5 of 5. The day a sequence check comes back as 4 of 5, you have a different kind of problem. Time drift means the standard is being stretched. Sequence drift means the standard is being replaced.

An audit that does not change a decision is theatre. This one supports a narrow set of practical choices.

You can decide where to look next. The result points at approve and OCR capture, and specifically at the approval step, which carries the largest single time gap. That is where a root cause conversation belongs. The comment Approver slow is a description, not a cause. Why the approver is slow is the question for the next analysis.

You can decide whether the standard time is still right. Three steps are hitting 100 per cent. That suggests the standard times for those steps are realistic. If a step never reaches its standard time across repeated audits, the standard may be wrong rather than the operator.

You can decide whether to escalate. With sequence intact at 5 of 5 and average adherence at 93.4 per cent, this is coaching territory, not a reopened project. The audit gives you the evidence to make that call calmly rather than by instinct.

The most frequent mistake is auditing from a desk. Observed time pulled from system logs misses queueing, rework loops and interruptions. The step that reads as 6.2 minutes in the sample is only visible as 6.2 minutes if somebody watched it.

The second mistake is treating the average as the answer. Average adherence of 93.4 per cent would let you close the file. The step detail would not. Always read the breakdown before you read the headline.

The third mistake is auditing once. A single audit gives you a position. Sustain needs a direction. Run the same five steps again next cycle and compare. Without a second audit you cannot tell whether 93.4 per cent is a decline from 100 or a recovery from lower.

The fourth mistake is making the audit feel like an inspection of people. Record behaviour against the standard, share the result with the team doing the work, and ask them what makes the standard hard to follow. Adherence that is enforced without explanation always decays again.

This tool measures conformance. It does not measure whether the standard work is any good. Every step could hit 100 per cent adherence and 13 minutes exactly while the process still produces the wrong outcome for the customer. Adherence to a poor standard is still a poor process.

It does not find causes. The result names approve at 81 per cent and carries the comment Approver slow. That is as far as it goes. Why the approval runs long, whether the approver is overloaded, whether the queue is upstream, whether the threshold for approval is set wrongly, none of that comes out of this audit. You need a separate analysis for that.

It is also a sample of one observation per step. Five steps observed means five observations. That tells you what happened during the audit. It does not tell you what happens on a Monday morning, during a month end peak, or with a relief operator. Treat one audit as a snapshot and build the picture across cycles.

Finally, the adherence percentage is a human judgement. The tool calculates faithfully from what you record. It cannot tell you whether you recorded it well. The discipline of honest observation sits with you, and it is the part of Sustain that no tool replaces.