Case: the "tickets closed" bonus

Reviewing an incentive starts with the number it was built to move, measured fairly: the same people, the same length of period, before and after. Only then look at everything it might have moved by accident.

At Klasna the bonus counts tickets closed. The data covers six coordinators for four weeks in March and four weeks in April. Before: 297 + 381 + 237 + 213 + 192 + 177 = 1,497 closes. Under the bonus: 335 + 479 + 238 + 257 + 245 + 220 = 1,774. Change: 277 / 1,497 ≈ +18.50%. On its own number, the bonus worked.

Say that plainly in the review. A proposal that hides the gain looks biased, and Iryna wanted more throughput in the peak for good reasons. The honest reading keeps the gain and then sets the costs beside it: the guardrails (reopens, satisfaction, hard tickets) come next.

Once the target's number is in, check the guardrail: the number that must not get worse while the target moves. For tickets closed, the natural guardrail is the reopen rate, because closing unsolved tickets shows up there.

Compute it the same way in both periods. Before the bonus, 84 of 1,497 closed tickets were reopened: 5.61%. Under the bonus, 16 + 73 + 4 + 16 + 29 + 20 = 158 of 1,774: 8.91%. Reopens grew 88.10% (84 → 158) while closes grew 18.50%: the side effect grew almost five times faster than the gain.

A guardrail that moves this much doesn't automatically cancel the target; it tells you the target is being reached partly by a shortcut. That's the question to put to the team and to Iryna: which part of the rise is real, and what change in the rule would keep the real part without paying for the shortcut?

When you measure a team, decide what each unit of the number stands for. A pooled rate divides all reopened tickets by all tickets closed, so every ticket counts once. An averaged rate takes each person's rate and averages them, so every person counts once, whatever their volume.

The two answer different questions. The pooled rate answers "what share of the tickets families got back were reopened?", which is what the bonus review needs. The mean of rates answers "how does a typical coordinator do?", and it lets a small queue weigh as much as a huge one.

At Klasna, Taras closed 479 tickets in April and Sofiia 220. Averaging their rates would give Taras's reopen rate of 15.24% the same weight as a much smaller queue. Pooling gives April's team rate: 158 reopened of 1,774 closed, 8.91%, against 84 of 1,497, 5.61%, in March. Say which rate you used in any report, so nobody compares one kind with the other.

A closed ticket is an event in the system; a solved problem is what the family wanted. Under a closes-only bonus the two drift apart, and the way to see it is to separate the extra closes into real new solutions and the same problems coming back.

At Klasna, closes rose by 277 under the bonus (1,497 → 1,774) and reopened tickets by 74 (84 → 158). So 74 / 277 ≈ 26.71% of the extra closes were the same problems closed a second time, each of them a family writing again. Tickets that stayed closed rose from 1,413 to 1,616: +203, or 14.37%, not 18.50%.

That's the honest size of the gain: real, smaller than the headline, and bought partly with families' patience. Whenever a count rises after an incentive, ask what share of the rise is genuine new work and what share is the incentive counting the same thing twice.

A good summary of an incentive puts the gain and every cost in one sentence, with numbers, and avoids both cheerleading and a takedown.

For Klasna's bonus weeks: closes rose 18.50%, but about a quarter of the extra closes were reopens; tickets solved without coming back rose 14.37%; satisfaction fell from 86.07% to 79.92%; and the coordinators solved 21 of the hardest tickets instead of 44, because an hour-long ticket paid the same as a two-minute one. Those hard tickets didn't disappear: they waited, or came up to the lead, the pattern from January all over again.

Two summaries to avoid. "The team became 18.50% more productive" counts reopens as output. "The bonus failed completely" ignores a real rise in solved tickets (1,413 → 1,616). And don't blame quality on who people are: the permanent staff moved too (Taras's reopens went from 9.45% to 15.24%). The cause is the rule, not the contract type.

An incentive's cost should be divided by the gain you actually wanted, not by the count it pays for. The rule says UAH 20 per ticket above 60, but the question for Iryna is what each extra solved ticket cost.

At Klasna the bonus paid UAH 7,200 in four weeks. Extra tickets that stayed closed: 1,616 − 1,413 = 203. So each extra solved ticket cost 7,200 / 203 ≈ UAH 35.47, before counting the cost of the hard tickets that waited (44 solved in March, 21 in April) and the drop in satisfaction.

Then look at the distribution. Taras earned UAH 4,780, 66.39% of the total, with the biggest rise in closes (381 → 479) and the worst quality (reopens 9.45% → 15.24%, satisfaction 67.14%). Maryna kept her pace and the team's best quality and earned UAH 60; Sofiia closed 43 more tickets and earned nothing. Report this to Iryna as what the rule paid for, not as a verdict on Taras: he did what the rule rewarded.

When a person games a badly designed rule, two things are true at once. The rule is at fault: Taras did what it paid for, so blaming his character would be unfair and would teach the team to hide shortcuts. And the effect is real: families wrote again because their tickets came back.

So the lead does both. The rule changes, for everyone, through the proposal to Iryna. The one-to-one with Taras uses the feedback shape from module 2: describe the behavior (closes 381 → 479, reopens 9.45% → 15.24%), its impact (families waiting twice for the same problem), and ask his view first before any next step. You might hear things the data can't show, such as which ticket types came back and why.

What to avoid: silence ("it was the rule, so nothing to discuss"), which leaves him unaware of the effect; and blame ("you gamed it"), which turns a design problem into a character judgement.

A better rule follows from what the review found. Test every option against the module's three lessons: does it pay for what you actually want, can you explain it to Maryna and to Taras in the same words, and does it fix a real cause?

Changes that address the side effects: count tickets that stay closed for 7 days instead of all closes, which removes the reward for closing twice; pay a team-level bonus only if the pooled reopen rate and satisfaction stay at March levels, which pays for the outcome and doesn't set people against each other; agree the new rule with the team before it starts, with a review at the end of May, which fixes the process Maryna named.

Changes that make it worse: a higher rate strengthens the same distortion, and a public ranking adds rivalry and gaming. Money never fixes Dmytro's refund skills and Bohdan's drafts. Recognition and a fair late-shift rota cost nothing.

How it fits together: read the target, then its guardrails, then who it paid; fix the rule, talk about the behavior. Next, module 5 turns to meetings and communication, starting with the meetings that eat the team's week.