Every improvement engagement ends with a promise that it will last. Very few end with a definition of when that promise has been broken. It is a comfortable arrangement for both parties: a promise that cannot be broken cannot be kept either, and then nobody has to defend anything.
Anchoring has become a closing courtesy. It appears in proposals and final reports, and it usually means little more than "we hope it sticks". For it to mean something, it has to point at something you can go and look at. And it can.
The half of PDCA that gets skipped
Plan-Do-Check-Act is so well established that most people can recite it. In practice the great majority of engagements stop after Do. The action has been carried out, the field is set to green, and the case is closed — closed because somebody said it was finished, not because anybody checked whether it worked.
It is worth noting that this is actually a requirement. ISO 9001:2015 states in clause 10.2.1 that the organisation shall review the effectiveness of any corrective action taken. The requirement is not that the action is carried out. The requirement is that somebody afterwards establishes whether it worked. In most management systems we meet, that particular clause has become a formality, filled in on the same day the action is closed — that is, before anyone could know.
You get what you measure
The usual measure is "actions closed on time". It is easy to extract and easy to understand, and it rewards precisely the wrong thing: closing. A team can hit a hundred per cent on-time closure and slowly get worse without it appearing anywhere. The number measures diligence, not effect.
It is the same mechanic that causes improvement projects to fade after three to six months: visible work gets recognised, and the work that prevents work from arising is invisible. The difference here is that one extra field fixes it.
Three things you can see at three months
Anchoring is not a feeling. It is three questions with observable answers, and anyone willing to take a walk and look in a log can ask them.
1. Has the deviation come back? Look in your own records for the three months that have passed. Not for the same wording, but for the same problem. "Medicine room not restocked" and "the medicine room was not filled up" are one deviation that happened twice, and a search on wording will not find it. If the problem has reappeared, the change did not hold, whatever the report says.
2. Does it hold without the person who introduced it? Find out who drove the change through, and look at what happened during the periods when they were away. Holiday, illness, a week on another project. If the routine collapses when one particular person is absent, it is not anchored in the system. It is anchored in a human being, and human beings change jobs.
3. Is it still being asked about? The third question is the one most often skipped, and it is the strongest. An organisation does not read what leadership said at the launch. It reads what leadership asks about three months later. If the measure has quietly dropped off the agenda and nobody noticed, the organisation already has its answer about how important it was.
Note that none of the three requires a new system, a consultant, or a measurement you do not already have. They require somebody to put an appointment in the calendar three months out and actually go and look.
What to do when the answer is no
What makes the three questions useful is not passing them. It is that a no points somewhere specific.
If the deviation has returned, you hit the symptom rather than the cause, and the task is a real root-cause analysis instead of a faster fix. If the routine collapsed without the key person, it has not been written down and handed on — it lives in someone’s head. If it has stopped being asked about, what is missing is neither a process nor a document. It is a management routine, and only management can build that.
Three different noes, three different tasks. That is why the questions are worth asking separately rather than collapsing them into one judgement about whether "it is running". And it is why we built the verification into our own tool for the daily operations meeting: when an action is closed with a real change, a date is set automatically fourteen days out, and the board asks once whether it held. Two buttons. It checks for itself whether the deviation has reappeared since the closure, and says so.
Fourteen days is not three months, and that is deliberate: the short loop catches the changes that never got going at all. The three months are the long loop, and they still require a human being to take a walk. You need both.
✓The main points
- →A promise of anchoring that cannot be broken cannot be kept either. Define in advance what you expect to be able to see in three months.
- →ISO 9001 clause 10.2.1 already requires the effectiveness of corrective actions to be reviewed. Most organisations fill in the field on the day the action closes — before anyone could know anything.
- →"Actions closed on time" measures diligence, not effect. A team can hit a hundred per cent and quietly get worse.
- →The three tests: has the deviation returned, does it hold without the person who introduced it, and is it still being asked about.
- →Each no points somewhere specific: root-cause analysis, documentation, or a management routine. Those are three different tasks.
References
- ISO 9001:2015, clause 10.2.1 — requirement to review the effectiveness of corrective action.
- Repenning, N. P. & Sterman, J. D. (2001). Nobody Ever Gets Credit for Fixing Problems That Never Happened. California Management Review, 43(4).
- Kotter, J. P. (1996). Leading Change. Harvard Business School Press — the eighth and final step is anchoring new approaches in the culture.
- Deming, W. E. (1986). Out of the Crisis. MIT Press.