Audits and inspections that find real problems

Audit versus inspection, and why it matters

The two words get used interchangeably and they are different activities with different purposes. An inspection looks at a physical thing at a point in time: is the guard fitted, is the extinguisher in date, is the scaffold tagged. An audit tests whether a process is working: were guards specified, installed, checked and maintained the way your system says they are.

The distinction matters because they find different problems. Inspections find the broken thing. Audits find the reason things keep breaking. An organisation that inspects heavily and audits lightly fixes the same issue repeatedly and never asks why it recurs.

• Inspections are frequent, short, and done by the people close to the work.
• Audits are less frequent, longer, and require independence from the area being examined.
• An inspection finding is usually a repair. An audit finding is usually a change to how something is decided or controlled.
• Both feed the same corrective action process — the difference is in what gets raised, not what happens afterwards.

The checklist problem

Most inspection forms have a structural flaw: they ask closed questions with an obvious right answer, in a fixed order, of someone who has answered them fifty times. The result is a form that can be completed accurately in ninety seconds from memory, without looking at anything.

This is not dishonesty. It is what happens when a form is designed to produce a compliance record rather than to direct attention. The give-away is the completion data: forms that are always fully compliant, always take the same time, and are always submitted at the end of a shift rather than during it.

Designing a form people answer honestly

Better forms share a small number of properties. They ask for observation rather than judgement, they require something that cannot be produced from memory, and they are short enough that the person filling them in has attention left for what they are looking at.

1. 1.Ask what was observed, not whether something is compliant. "How many of the six extinguishers were in date?" beats "Are extinguishers compliant?"
2. 2.Require a photograph on the items that matter. A photo cannot be recalled from memory and takes three seconds.
3. 3.Randomise or rotate the order of sections, so the walk changes and attention does not settle into a route.
4. 4.Cut the form. Twelve questions answered properly beat forty answered from habit. Move the rest to a quarterly deeper check.
5. 5.Make raising a problem the easy path, not the one that generates work for the person raising it. If flagging an issue means filling in three more screens, people stop flagging issues.

Sampling that holds up

You cannot audit everything, and auditors do not expect you to. What they expect is that your sample was chosen for a reason you can state. A defensible sample is not necessarily a large one — it is one where the selection method is deliberate and recorded.

• Sample across the risk, not evenly. The high-consequence process gets looked at more often than the low one.
• Include the awkward cases deliberately: night shift, the remote site, the new contractor, the process that changed last month.
• Follow single items all the way through rather than checking many items shallowly. One permit traced from request to close-out finds more than twenty permits glanced at.
• Record why you chose what you chose. "Selected the three highest-risk tasks performed since the last audit" is a defensible statement; a list with no rationale is not.

The same logic applies to inspections. Rotating which areas get inspected each week, weighted by risk and by where problems were found last time, beats a fixed rota that everyone can predict.

Findings that get fixed

A finding is a piece of writing whose only purpose is to cause a change. Most findings fail at that, and they fail in predictable ways: they describe a symptom without evidence, they name a person, or they prescribe a solution the writer preferred instead of stating the problem.

1. 1.State the requirement, the evidence and the gap, in that order. Requirement first makes the finding about the system rather than the person.
2. 2.Quote what you saw and where. Specificity is what makes a finding actionable and what stops it being argued away.
3. 3.Do not write the fix into the finding. The person who owns the process should choose the control; your job is to define what has to be true afterwards.
4. 4.Grade honestly. Inflating a minor issue to force attention destroys the grading system for the finding that genuinely warrants it.
5. 5.Give it an owner and a date at the moment it is raised, not in a separate meeting a fortnight later.

Then track them in one place. The most common audit finding about audit programmes is that previous findings were never closed — an entirely self-inflicted wound, and the easiest one to avoid.

Questions we get asked