Layered Process Audit: How to Tell If Your Improvement Is Quietly Slipping

The Layered Process Audit tool lets you define audits conducted at multiple organisational levels to verify process adherence and control health. One layer is not enough. A supervisor checking the same five things every day will eventually stop seeing them. A director visiting once a month will see the room with fresh eyes but will miss the daily detail. Layering the checks covers both.

You reach for this tool in the Sustain phase of RDMAICS, after Control has produced a control plan and after handover to the process owner. Control tells you what should be monitored. Sustain tells you whether anybody is actually doing it, month after month.

The output is deliberately simple. You get a list of layers, who owns each one, what it looks at, how often it runs, how many items get sampled and the last score recorded. That is enough to see whether your audit system is healthy, and enough to see when it is not.

Everything below uses the sample data that ships with the tool. It is fictional. No real site, team or result is described here. The value of the example is that the numbers are close enough to a real audit programme to show you how to read one.

The case has four layers. Layer 1 is owned by a supervisor, focused on daily adherence, run daily, with a sample of 5 items, last score 95%. Layer 2 is owned by a manager, focused on process and board, run weekly, with a sample of 10 items, last score 90%. Layer 3 is owned by a director, focused on system and governance, run monthly, with a sample of 20 items, last score 92%. Layer 4 is owned by a Black Belt, focused on control plan and SPC, run monthly, with a sample of 15 items, last score 94%.

Read that structure before you read the scores. The frequency drops as you move up. The sample size rises, except at Layer 4. The focus widens from one shift to the whole governance system. That shape is the point of a layered audit.

The table takes seven columns. Audit layer, auditor level and last score are required. Focus, frequency and sample size are optional but you should fill them, because without them the scores float free of any context.

Name the layers in order of seniority, not in order of importance. Layer 1 sits closest to the work. Layer 4 sits furthest away. The auditor level column names the role that owns the layer, not the individual. Roles survive staff changes. Names do not.

Keep the focus field short and specific. In the sample data, the four focus statements are daily adherence, process and board, system and governance, and control plan and SPC. Each one is a different question asked of the same process. None of them overlap. If two of your layers describe the same focus, you have four audits but only three layers of coverage.

Frequency and sample size set the cost of the programme. Daily at 5 items is cheap. Monthly at 20 items is not cheap, but it runs twelve times a year rather than two hundred and fifty. Balance them so the total effort is something the organisation will still be doing in six months.

The tool calculates four headline figures from the sample data. Layers: 4. Average last score: 92.8%. Lowest: Layer 2, Manager, at 90%. Highest: Layer 1, Supervisor, at 95%. The range runs from 90% to 95%.

Start with the range, not the average. A five point spread across four layers is tight. That tightness is itself a finding. It says the different levels of the organisation are broadly agreeing about what they see, which is what you want from a mature audit programme.

Then look at where the low score sits. In this case the lowest is Layer 2, the manager layer, focused on process and board at a weekly cadence with a sample of 10 items. The highest is Layer 1, the supervisor layer, focused on daily adherence with a sample of 5 items. A higher daily score and a lower weekly score is a familiar pattern. The thing checked every day gets done. The thing checked weekly, which here includes the visual board, slips a little between checks.

Each layer also carries a checked state. In the sample all four read Not checked. That is the tool telling you the audits are defined but not yet marked complete for the current cycle. An AI review is available if you want a written explanation of what the scores imply.

A layered audit is not a scoreboard. It is a decision aid. Here is what the sample result lets you decide.

The most common decision is where to spend attention next. With Layer 2 at 90% and Layer 1 at 95%, attention goes to the process and board focus at the weekly cadence, not to daily adherence. Daily adherence is holding.

Slippage rarely announces itself with a bad score. It announces itself with a missing one. The first sign is a layer that stays in the Not checked state past its due cycle. In the sample data, all four layers read Not checked. On day one that is normal. Three weeks into a daily audit it is not.

The second sign is divergence. If the gap between the highest and lowest layer widens well beyond the five points seen here, the levels of the organisation are no longer seeing the same process. Usually that means the layer closest to the work has started scoring the audit rather than the process.

The third sign is the upper layers going quiet. Layer 3 and Layer 4 both run monthly in the sample. Monthly audits are the easiest to postpone, because nothing visibly breaks when you skip one. Two skipped months and the governance layer has effectively been removed.

Watch the direction of travel as much as the level. A layer that moves from 95% to 92% to 90% over three cycles is a clearer warning than a layer that sits flat at 90%.

The first mistake is stacking layers that all ask the same question. Four auditors checking daily adherence is one audit done four times. The sample data avoids this by giving each layer a distinct focus.

The second is oversampling at the top. Layer 3 samples 20 items monthly and Layer 4 samples 15. Those are the largest samples in the programme and they belong at the layers that run least often. Reversing that, with large samples at a daily cadence, is how audit programmes die of exhaustion.

The third is treating the average as the result. The average last score of 92.8% is a summary, not a finding. Nothing in the sample data actually scores 92.8%. The findings live in the individual layers and in the gap between them.

The fourth is leaving the auditor level blank or filling it with a person's name. Roles persist. If the manager who owns Layer 2 moves on, the layer should move with the role, not disappear.

This tool records a schedule and a last score. It does not run the audit. Nobody is checking anything because a row exists in a table. The discipline sits with the owners named in the auditor level column.

It does not store history. You get the last score for each layer, not a trend. The sample result tells you Layer 2 is currently the lowest at 90%, but it cannot tell you whether that is an improvement or a decline. If you want direction of travel, you need to keep your own record of previous cycles.

It does not weight the layers. The average of 92.8% treats a daily supervisor audit of 5 items and a monthly director audit of 20 items as equally important. They are not. Read the layers individually before you read the average.

It does not verify the scores. A layer can read 95% because the process is sound or because the audit is being completed at a desk. The tool has no way to tell those apart. That is why the layering matters. A senior layer that consistently scores lower than the layer beneath it is the only cross check you have, and it is a check on the auditors as much as on the process.

Finally, it does not tell you what to do about a low score. It tells you where the low score is. In the sample data that is Layer 2, Manager, at 90%, focused on process and board. What to do about that is a conversation with the owner, not an output of a tool.