Goals people accept

Research on goal setting agrees on four conditions. A goal works when it is specific (a number and a date, so people know where they stand), hard but reachable (a stretch the team believes in), accepted by the people who carry it, and followed by feedback on progress.

Each missing condition breaks the goal in its own way. "Let's all do our best in May" is kind and vague, so nobody can tell whether they're on track. "Close 20% more tickets, Iryna has decided" is specific but imposed, says nothing about quality and gets compliance at best. "Zero reopened tickets" is specific and out of reach, so people drop it.

A goal that meets all four for Klasna's Support in May: first reply within 45 minutes on 90% of new messages on shift, with the reopen rate no higher than in March, numbers agreed with the team and checked every Monday. Note the second number: a target with a guardrail beside it.

"Do your best" sounds like the kind choice for a tired team. In practice it works worse than a specific, hard but reachable goal, and this is one of the most consistent findings about goals.

The reason is simple: with "do your best" people can't see whether they're on track, so they quietly settle lower than they could reach. A clear target with visible progress gives direction and usually feels better than vague pressure, because people know when the week went well.

Hard but reachable is the balance. Too easy and the goal adds nothing; out of reach and people stop trying. Tiredness changes the number, not the need for one: in the middle of Klasna's exam peak, agree a reachable target with the team and protect their capacity (fewer meetings, cover for reviews) instead of dropping the target and hoping for the best.

A goal people accept pulls effort; a goal handed down without a reason gets compliance: people do the minimum the number demands and stop caring about what it was for. The difference is commitment, and you can build it.

Two moves do most of the work. Explain why the goal matters now, with the facts behind it. Then let the people who carry the goal propose the number, or at least test yours against their experience. People who helped set a target understand it better and defend it when the week gets hard.

At Klasna, a May goal on first replies works better if you show the team April's numbers, say why families notice the first reply in the peak, and ask the team what is hard but reachable on shift. You still keep the final say if the proposal is far too low or too high: agreed doesn't mean anything goes. And a stretch must stay believable; an impossible number gets dropped, not half-met.

A goal without feedback is a wish: people can't tell whether they're on track, so effort drifts. The fourth condition of a working goal is regular, visible feedback on progress.

Keep it simple and frequent. A short weekly look at the shared number, at team level, with a sentence on what helped and what got in the way. In a distributed team like Klasna's Support (Kyiv, Lviv, two people abroad), write it down where everyone can see it, because remote people don't overhear how the week went.

Feedback on a team goal is not a public league table. Ranking each coordinator's numbers every week turns a shared goal into a contest, pushes people to protect their own count instead of helping each other, and punishes careful people like Maryna, whose numbers are lower and whose replies are the best. Individual numbers belong in one-to-ones, read together with quality, not on a wall.

Specific, hard targets have one important exception: work that is new and complex for the person. While the skill isn't there yet, a performance number (how many, how fast) pushes people to rush and guess instead of learning.

For that work, set a learning goal first: what the person will be able to handle, by when, and with what support. Dmytro is new to refund disputes, which are rule-heavy. "Close 40 disputes in May" would teach him to guess faster. "By the end of May, handle the five common exception types, with Olha reviewing your first ten cases" builds the skill. The volume or speed target can come once he handles the cases confidently.

A learning goal needs real support, and support costs someone time. Olha agrees to review ten cases if it goes into her week, not on top of it, so the lead covers two hours of her queue. Budget the coaching, or the learning goal quietly becomes extra unpaid work for the reviewer.

A bonus is a goal with money on it, and Klasna's April bonus counts one thing: tickets closed above 60 a week. When one number decides the money, people steer towards the number, and the work it doesn't count gets squeezed.

You can predict the side effects by asking what raises the count cheaply. Closing a ticket before the problem is fully solved. Leaving long, hard tickets for someone else, because an hour-long ticket pays the same as a two-minute one. Splitting one family's two questions into two tickets. A reopened ticket closed again even counts twice.

You can also predict what gets discouraged: anything that costs time and adds nothing to your own count, such as checking that a refund really went through or helping a colleague with a tricky ticket. That work protects quality and the team, and a closes-only rule makes it quietly irrational. Read any target this way before it starts: what does it pay for, and what does it stop paying for?

When one number decides the money, people improve the number, and sometimes that drifts away from what the number was meant to show. This is metric gaming, and it is rarely cheating: people respond sensibly to the rule they were given.

So read gaming as a fault in the measure, not in the person. Under Klasna's April bonus, a reopened ticket closed again counts twice, and a family's ticket was closed twice without a refund. Taras broke no rule; the rule paid for closing, not for solving. Iryna owns the design ("I wanted speed in the peak") and asks for numbers and a proposal.

The fix is usually one of three: count something closer to the real result (tickets solved without a reopen), pair the target with a guardrail that must not get worse, or keep the number for learning and take the money off it. Dropping every measure is the opposite mistake: then nobody can see progress at all.

You don't have to drop targets to avoid gaming. Pair each target with a guardrail: a second number that must not get worse while the target moves. The same idea protects product experiments in the course Product Analytics: From Data to Decisions.

Choose the guardrail by asking which shortcut the target invites. Tickets closed per week pairs with the share reopened, because closing unsolved tickets shows up as reopens. First reply within 45 minutes pairs with satisfaction, because a fast but empty reply shows up there. Replies started from an AI draft pair with reopens on those tickets. Fewer hand-ups is good only if hard tickets don't simply sit in the queue.

Read targets and guardrails at team level, so the guardrail doesn't become a new league table. And when a guardrail moves the wrong way, talk with the team before any money changes: it's a question, not an automatic penalty.

How it fits together: specific, hard, agreed goals with feedback pull effort; a learning goal fits new work; every target needs a guardrail. Next, Fairness and recognition: why the same bonus feels so different to Taras and to Maryna.