Public protocol · version 1.0

Everyone promises the improvement will last. Here is how you check it.

Five checks, 90 days after a change was closed. They take an hour and a half, they need no access to your data, and you can run them without us — including on work we did not do.

A promise that cannot go wrong is not a promise

Every adviser in the country writes that the improvements last. None of them writes how you would see whether that was true. So the sentence cannot be disproved — and a sentence that cannot be disproved does not separate the capable from the articulate.

The usual answer is testimonials and case studies with percentages. They do not work as evidence, and the reason is simple: nobody publishes the case that went wrong. The reader knows it, and so they do not count.

This protocol costs more to publish. It is written before the result is known, it is dated and versioned, and it is worded so that you can run it yourselves. That means it can come out against us. That is why it is worth reading.

The five checks

Each check is something you can walk over and look at. None of them is a judgement call, and none of them can be answered from a desk. The order is not accidental: the first four measure whether the change is alive, and the fifth measures whether it has been tested.

1

Is the change still there?

Can you walk over and point at it?

How you do it

Go physically to where the change was made. Find the standard, the board, the form or the routine. Look at when it was last written in, or when it was last followed.

It passes when

It exists on site, and its most recent use falls within the interval it prescribes for itself. A daily routine last used nine days ago is not a pass.

Why the check is there

The most common outcome is not that the change was rejected. It is that it quietly stopped being used, without anyone deciding so. That can only be seen on site — never in a status report.

2

Does anyone besides its author use it?

Can two people who were not involved explain the routine the same way?

How you do it

Ask two people who did not take part in the work how the task is carried out. Let them explain without looking anything up. Compare the two answers with each other — not with the standard.

It passes when

Both describe the same thing, and the description matches the standard. If the two answers differ, the routine is not shared, whatever the paperwork says.

Why the check is there

An improvement that lives only in the head of the person who made it is a dependency on a person, not a system. It leaves when they change jobs — and it is the most common reason an operation stalls six months after someone resigned.

3

Has the number followed?

Has the measurement moved — and is the movement signal or noise?

How you do it

Take the key figure the change was meant to move. Compare the last 30 days with the 30 days before the change. Judge the difference against the process’s own spread, not against an average: a shift counts only once it falls outside what the process usually swings.

It passes when

A real signal in the right direction. Or: a stable process where the deviation has stopped appearing. “It looks a bit better” is not a pass.

Why the check is there

Almost all reporting treats any movement as an event. If scrap went from 2.1 to 2.6, that was usually noise — and adjusting the process because of noise makes it worse. Deming called it tampering. An anchoring check that does not tell the two apart is itself a source of error.

4

Has the problem stopped coming back?

How many times has the same problem been seen since?

How you do it

Count how many times the original problem was recorded in the 90 days after closure, and compare with the 90 days before. Count PROBLEMS, not records: the same problem described in five different wordings is one relapse.

It passes when

Fewer than before, and none in the last 30 days.

Why the check is there

It is the only one of the five that measures the root cause. The others measure whether the solution is alive; this one measures whether it was the right one. A routine can run impeccably and still be the answer to the wrong question.

5

Has it survived a disruption?

Has anything happened since you closed it — and did the routine hold?

How you do it

Find out whether the 90 days held a peak load, a holiday period, an absence in a key role, a new employee or a reorganisation. Then look at whether the routine ran through that period.

It passes when

At least one disruption has occurred, and the routine ran through it. If nothing has happened, the check is NOT passed — it is unresolved, and it must be recorded as unresolved.

Why the check is there

This is the check nobody thinks of, and it is the most important one. A routine that has only run in calm weather is untested. Everything holds until something happens. So an improvement can pass the first four checks and still not be anchored — it just has not been out in the weather yet.

What the outcome means

No points are given for a partial pass. A check is either passed or not passed, and one unanswered check counts as not passed. The reason is that a scale always ends up negotiated — and a yardstick that can be negotiated measures nothing.

5 of 5

Anchored

The change is alive, it is shared, the number followed, the problem is gone, and it has stood through a disruption. It is the only outcome that earns the word.

3-4 of 5

Fragile

Some of it holds, some of it does not. The outcome is only useful if you write down WHICH check failed — each of the five points to its own cause and its own next step.

0-2 of 5

Not anchored

The change lived on attention and left with it. That is not a failure, it is a measurement. What was wasted is not the effort — it is the months in which nobody knew.

Why 90 days

Three things have to be over before the answer means anything: the attention from the project, the first busy period, and the first absence in a key role.

Measure after a fortnight and you measure whether people can remember something. Measure after a year and the traces are gone, and nobody can say why. 90 days is the earliest point at which the answer is worth having, and the latest at which the cause can still be seen.

How it relates to ISO 9001

ISO 9001:2015 clause 10.2.1 requires the effectiveness of a corrective action to be reviewed. The requirement is there. How you do it is not.

So in most management systems the field is filled in the same day the action is closed — which is logically impossible, because the effect has not happened yet. The protocol is a concrete answer to that requirement, not an addition to it.

What we commit to

  • The protocol is run on every engagement we close — including when we have a feeling about how it went.
  • The result is yours, in writing, whatever the outcome. If checks failed, it says which.
  • If we change the protocol, it gets a new version number and a date. The edition your engagement was measured against stays as it is.

Version 1.0, published 21 September 2026. Free to use, including without us, including by other advisers.

Download the check sheet (in Danish)Write to us

The terms used in the protocol