Measures change what they measure
A measure that is only observed usually describes the work. A measure that is attached to a target, a bonus, a ranking or a reputation starts to shape the work. People are not being dishonest when this happens; they are responding sensibly to what the organization appears to value.
A popular way of putting it is that when a measure becomes a target, it ceases to be a good measure. The saying is loosely worded, but the underlying point is widely recognized: pressure to hit a number creates incentives to hit the number, not always to achieve what the number was meant to show.
Common patterns
The examples below are illustrative, not drawn from any particular organization.
Narrowing focus. A team measured on speed neglects accuracy. A team measured on cost neglects service quality or future maintenance. Whatever is not measured quietly receives less attention.
Choosing easy work. If the target is the number of items closed, simple items get handled first and difficult ones are left to age. The count looks healthy while the hardest problems pile up.
Redefining or reclassifying. A ticket is relabelled as a different type so it falls outside the measured category. A clock is paused for a questionable reason. The measure stays on target; its meaning drifts.
Splitting and bundling. One piece of work is split into several to raise the count, or combined to hide a late item.
Timing effects. Work is rushed or held back around the end of a period so that the numbers fall on the right side of a line. Results then bunch up just before deadlines.
Short-term over long-term. A cost target may lead to postponed maintenance or training. The saving appears now and the cost arrives later, often under someone else’s budget.
Discouraging bad news. If reporting a problem is penalized, problems are reported late or not at all. A reassuring dashboard can then reflect a reluctance to speak up as much as real performance.
Gaming versus honest adaptation. Not all response is manipulation. People may also reasonably prioritize in line with the target. The distinction matters because the remedy differs: punishing honest adaptation tends to make matters worse, while fixing the target changes the incentives.
Why centralized settings are exposed
Central functions often serve many internal customers and are judged on a handful of summary numbers. Standard targets set at the center may not suit every site or business unit. A uniform deadline can be easy in one location and unrealistic in another, which can push local teams toward workarounds, or lead them to feel the measure is imposed on them. Resistance to targets may overlap with broader change resistance, and it is a signal worth listening to rather than overriding.
Designing measures that hold up better
No design is immune, but several habits reduce the problem.
Use a balanced set. Pair measures that pull in different directions, such as speed with quality, or cost with satisfaction. If gaming one harms another, the distortion shows up.
Measure outcomes as well as activity. Counting actions (calls made, documents produced) is easy to inflate. Outcomes closer to what the organization wants are harder to fake, though often slower and noisier.
Keep definitions tight. Clear written definitions, including rules about reclassification and pausing clocks, leave less room for drift. Metric definitions need an owner, and changes should be logged.
Check quality by sampling. Periodically examine a sample of real cases, whatever the numbers say. A handful of reviewed tickets or invoices can reveal practices that the aggregate hides.
Use ranges rather than single hard thresholds. A target of “between X and Y” or a trend direction can reduce end-of-period cliffs better than a pass-or-fail line. Be thoughtful about whether any target is needed at all; sometimes simply watching a measure is enough.
Set targets with the people who do the work. Staff usually know where a number can be gamed and where it is unrealistic. Involving them improves the target and the buy-in. Be open about how the data will be used.
Review targets regularly. A target that fitted last year may no longer fit. Retire or adjust measures when they begin to distort, and look at governance reviews to ensure this happens.
Look for unusual patterns. Clusters of results just inside a threshold, sudden changes in classification, or a measure improving sharply with no other signs of change are prompts for questions, not proof of misconduct. An audit trail on key data helps answer them.
Separate learning from judging. If measures are used mainly to understand and improve, people share information more freely than when every number is tied to personal consequences. Where pay or ranking is linked to a figure, the risk rises and caution is warranted.
A short illustration
Imagine a support team told to cut average resolution time. Over a quarter, the average falls. A closer look finds three things: some tickets were closed as soon as an automated message was sent, a share were reopened later, and satisfaction comments mention having to repeat information. The target was met and the service worsened.
A better set might combine resolution time with reopen rate and a satisfaction score, review a sample of closed tickets each month, and define clearly when the clock may pause. The detail of those measures is covered in the article on service desk metrics.
Know the limits
These safeguards lower the risk; they do not remove it. Any measurement system involves judgement and trade-offs, and sometimes the most honest answer is to accept a measure’s imperfection and use it alongside conversation and observation. Targets can still be helpful for focusing attention, as long as the organization stays alert to what they may be doing to behavior.