You want to improve something in your operation, so you start measuring it. That is sensible, and it is what every management book tells you to do. But there is a catch that those books mention less often, and it is the single most important thing to understand about metrics. The moment you choose a number to track, you have not just started observing your team. You have handed them an instruction. What you measure tells people what to optimise, whether you meant it that way or not, and they will optimise it. The behaviour follows the measure, reliably and fast. The only real question is whether the thing you actually cared about comes along for the ride.
Often it does not. You pick a metric as a convenient stand-in for something you care about, because the real thing is hard to measure and the stand-in is easy. Then the stand-in becomes the target, the team optimises the stand-in, and the gap between the number and the thing you actually wanted quietly widens until the metric is going up while the reality goes down.
The metric becomes the target
The classic example lives in customer support. You want happier customers, but happiness is hard to measure, so you measure something adjacent and countable: tickets closed per day. It seems reasonable. More tickets closed sounds like more problems solved. But you have just told your team that the goal is closing tickets, and closing tickets is not the same as solving problems. The fastest way to close a ticket is not always to fix the issue; it is to reply quickly, mark it resolved, and move on. So tickets get closed fast, the metric climbs, and everyone looks productive. Meanwhile customers whose problems were not actually fixed reopen those tickets or, worse, quietly leave. The number you are watching says things are improving. The thing you cared about is getting worse.
This is not a failure of the people. It is a completely rational response to the instruction you gave them. They are doing exactly what you measured. The failure was in choosing a measure that could be satisfied without achieving the goal.
Every metric has a shadow
Every metric you pick encourages something you did not explicitly ask for, and it is worth naming that shadow before you commit to it. Measure call handling time and you encourage agents to rush people off the phone. Measure lines of code and you encourage bloated, padded code. Measure utilisation and you encourage people to look busy rather than to think. Measure sales calls made and you encourage a flurry of low-quality calls to no one in particular. In each case the metric captures a fragment of what you want and leaves out the rest, and the part it leaves out is exactly where the behaviour quietly degrades.
The trouble is that the shadow is invisible on the dashboard. The dashboard shows the metric going the right way. It does not show the customer who was rushed, the code that will need rewriting, or the deals that were never going to close. So the very tool you introduced to give you visibility can end up hiding the thing you most needed to see.
Why the wrong metric is worse than none
It is tempting to think a flawed metric is at least better than flying blind. Usually it is not. With no metric, at least people use their judgment about the actual goal. With the wrong metric, you have actively pointed that judgment in the wrong direction and attached consequences to it. You have not just failed to measure the right thing; you have paid your team, in attention and sometimes in bonuses, to optimise the wrong thing. And because the number moves in the right direction, everyone believes it is working, which means the problem can run for a long time before anyone notices that effort and outcome have come apart.
Measuring the right thing
The fix is not to abandon measurement. It is to measure with more care about what the number will actually make people do. Start from the outcome you genuinely want and choose a measure as close to it as you can bear, even if that is harder than counting the easy thing. Prefer outcomes over activity: problems resolved rather than tickets touched, orders delivered on time rather than orders processed. Where any single number can be gamed, pair it with a balancing one, so that winning on the first while losing on the real goal shows up immediately in the second. Speed of response is fine as long as it sits next to reopen rate. Volume is fine as long as it sits next to quality.
Above all, watch the behaviour your metric produces, not just the figure it reports. If the number is improving but the people closest to the work seem to be doing something that feels wrong, believe the behaviour over the dashboard. The behaviour is telling you the truth about what you are really measuring.
A worked example
A company asked us to help because their support team's numbers looked excellent but their customers were unhappy and churn was creeping up. The team was measured, and bonused, on average ticket resolution time, and they had driven it impressively low. On paper, support was a triumph. In reality, agents had learned that the way to a fast resolution time was to close tickets quickly, so they were closing them the moment they had sent any reply at all, whether or not the customer's problem was solved. Customers reopened tickets, which technically counted as new tickets, which the team also closed quickly. The metric looked better and better as the service got worse and worse.
We did not lecture the team, because the team was behaving perfectly rationally. We changed what was measured. Resolution time stayed, but it was paired with a reopen rate and a simple first-contact-resolution measure, so closing a ticket that was not really solved now hurt the numbers instead of helping them. Almost immediately the behaviour changed, because the instruction had changed. Agents spent a little longer per ticket, closed fewer of them in a rush, and the reopen rate and the churn both fell. The team was no less capable than before. They had simply been aimed at the wrong target, and re-aiming them was most of the work.
Measure what you actually want
Before you put a number on a wall and ask a team to move it, ask one question: if someone optimised this metric perfectly while ignoring everything else, would I actually get what I want? If the answer is no, you have found a metric that will mislead you, and the more successfully your team hits it, the more misled you will be. Choose measures that only move when the real thing moves, pair them so they cannot be gamed in isolation, and keep watching the behaviour underneath the number.
Getting the right things measured, so your operational data points at reality rather than away from it, is exactly what our reporting and dashboards work is built around: turning numbers into a true picture of how the operation is actually performing. Book a discovery call and we will help you measure what you actually want.
Frequently asked questions
What does "you get the behaviour you measure" mean?
It means the metric you choose to track is effectively an instruction to your team about what to optimise, whether you intend it that way or not. People naturally steer towards the number they are judged on, so if you measure how many tickets are closed, you get more closed tickets, and if you measure call handling time, you get shorter calls. The behaviour follows the measure. The important consequence is that this happens even when the metric is a poor stand-in for what you actually care about, so a badly chosen metric quietly pulls the whole team towards the wrong goal.
Why can a metric make performance worse?
Because once a number becomes the target, people optimise the number rather than the outcome it was meant to represent, and the two can diverge. If support is measured on tickets closed per day, the fastest way to score well is to close tickets quickly rather than to actually resolve the customer's problem, so tickets get closed and reopened, and customers get more frustrated even as the metric improves. The metric goes up while the real performance goes down. A wrong metric is worse than no metric, because it actively directs effort towards the wrong behaviour while feeling like progress.
How do I choose the right operational metric?
Start from the outcome you actually want, not the activity that is easiest to count, and choose a measure as close to that outcome as you can get. Prefer measures of results, such as problems genuinely resolved, over measures of activity, such as tickets touched. Where a single metric can be gamed, pair it with a balancing one so that improving the first at the expense of the real goal shows up in the second, for example speed alongside reopen rate. And watch the behaviour the metric produces, not just the figure, because the team's response tells you whether you are measuring the right thing.
Should we measure activity or outcomes?
Outcomes, wherever you can, because activity is easy to count but easy to inflate without creating any value. Counting calls made, tickets touched, or meetings held rewards motion rather than results, and a team optimising an activity metric can look extremely busy while achieving very little. Outcome measures, such as issues resolved first time, revenue collected, or orders delivered on time, are harder to define but far more honest, because they only move when something real happens. Activity metrics have their place as leading indicators, but they should never be the thing you ultimately judge success by.