- Assume every measure will be gamed and design so that gaming it produces the outcome you wanted.
- Pair every speed measure with a quality measure, or you will get speed at the cost of everything else.
- Require evidence for the lines that are easiest to claim and hardest to verify.
Gaming is a design problem, not a character problem
When a measure affects pay, promotion or standing, people optimise it. This is not a moral failing, it is exactly what you asked for by measuring it. The useful question is never whether people will optimise, it is what the cheapest route to a good score looks like.
If the cheapest route is doing the job properly, the scorecard is well designed. If the cheapest route is ticking boxes in the car park, no amount of reminding people to be honest will fix it.
Pair every measure with its opposite
Measure service speed alone and you get rushed service. Measure quality alone and you get slow service. Measure availability alone and you get overstocking and waste.
Every measure has a natural failure mode, and the fix is to pair it with the thing it would otherwise destroy. Speed with accuracy, availability with waste, audit score with mystery visit score.
Work out the laziest way to score well on your scorecard. If that route also runs a good site, you have designed it correctly. If not, you have designed a form-filling exercise.
Require evidence where claims are cheap
Lines like "washrooms checked hourly" are free to claim and expensive to verify. Those are exactly the lines that need a timestamped photo rather than a tick.
You do not need evidence on everything. You need it on the small number of lines where the gap between claiming and doing is widest.
Watch for suspiciously perfect scores
A site returning 100 percent every week is not a star, it is a flag. Real operations have bad days, and a scorecard that never records one is measuring the recording rather than the operation.
Treat perfection as something to go and look at, in the same way you would treat a sudden decline.
Frequently asked questions
Should scorecards be linked to bonuses?
They can be, but only once the scoring is consistent and evidence-backed. Linking pay to a noisy measure teaches people to attack the measurement.
How many measures should a site scorecard have?
Five to eight. Beyond that nobody can hold it in their head, and measures nobody remembers do not change behaviour.
What do I do if I catch a site gaming the scorecard?
Fix the scorecard first. If one site found the gap, others have too, and punishing the one that was caught leaves the design fault in place.