The Measure That Changes the Work
Why any measure people are judged on reshapes the work it measures, the four forms that reshaping takes, and how to design a measure that survives being known.
Any measure that people are judged on will change the work, and the change is usually invisible in the measure itself. Count calls and calls get shorter. Count closures and difficult cases get reopened under a new reference. Count activity and activity appears.
The measurement limits in “The Measure That Changes the Work” provide useful context for what teams should check in monitask pricing when researching monitask pricing. Time and activity data can reveal a workflow question, while outcomes, employee explanation and a documented human review remain necessary to understand why a pattern appeared and whether any action is justified.
None of that requires dishonesty. It is what happens when people are rational about what they are assessed on, and it is entirely predictable at the point the measure is chosen.
Teams reviewing “The Measure That Changes the Work” can cross-check their approach against the Massachusetts wage and hour laws. The comparison is most useful when the organisation records which recommendations apply, which do not and why.
The four forms it takes
Displacement: effort moves from unmeasured work to measured work. The unmeasured work was often the valuable part.
Redefinition: the unit shifts. A case becomes three cases; a call becomes two calls.
Timing: work is moved between periods to smooth or inflate a figure.
Minimum compliance: people do exactly what the measure requires and stop, which is the most damaging and the hardest to see.
Why this is not cheating
It is worth being explicit about internally, because the instinct is to treat the response as misconduct.
An employee who shortens calls because call length is measured is doing what the organisation asked. The problem is the measure, and treating it as a conduct problem both misdiagnoses it and teaches people to hide the adaptation rather than to mention it.
Designing a measure that survives
- Name what the measure is a proxy for, in one sentence.
- Ask how somebody could satisfy the measure without producing the thing.
- Add a counterweight measure that gets worse if they do.
- Keep both visible, always together.
- Review whether the relationship still holds, annually.
- Change it when it stops holding, rather than policing the response.
Step three is the whole technique. Volume with quality. Speed with rework. Activity with output. A single measure is always gameable; a pair that pull against each other is much less so.
The measures in this collection
Idle time can be satisfied by moving a mouse. Activity scores can be satisfied by having the right application in focus. Punch records can be satisfied by somebody else's hand.
Each of those is the predictable response to measuring an input, and each is cited as evidence of dishonesty when it is mostly evidence of a badly chosen proxy.
Measuring quietly does not help
The apparent solution — do not tell people what is measured — fails on its own terms and creates a worse problem.
It fails because people work it out. And it creates a worse problem because an undisclosed measure is both harder to justify and more damaging when discovered, for no gain over a disclosed one with a counterweight.
What the adaptation tells you
The way people adapt is an unusually honest description of the organisation's priorities as they are actually experienced.
If measuring closures produces reopened cases, the organisation has told people that closures matter more than resolution. That is useful information about the measure and about everything else that is signalled alongside it.
Reviewing a measure
Once a year: what is this a proxy for, does the relationship still hold, what has changed in how people work, and what is the counterweight.
Four questions. Measures are introduced with care and then run for a decade through system changes, role changes and reorganisations that quietly break the relationship they depended on — and nothing in the reporting ever says so.
Watching for the adaptation deliberately
Having predicted how a measure could be satisfied without producing the thing, look for exactly that within the first quarter.
- Has the unit of work changed in size or definition?
- Has anything moved between periods near a boundary?
- Has unmeasured work visibly declined?
- Has the counterweight measure moved in the wrong direction?
Four questions, asked once, three months in. They catch the adaptation while it is still a design question rather than an established practice.
Telling people you are watching for it
Said openly — we expect this measure to change behaviour and here is the counterweight we are watching — it becomes a shared design problem rather than a game.
People will then tell you how the measure is distorting their work, which is information no amount of analysis produces, and which is the fastest route to a better measure.
The connection to this collection
Every control described on this site changes the behaviour it measures, including the ones that work.
That is not a reason to avoid controls. It is a reason to predict the adaptation before deploying, to measure whether it happened, and to treat what you find as information about the design rather than as evidence about the people.