"When a measure becomes a target, it ceases to be a good measure."
Goodhart's Law, in Marilyn Strathern's phrasing

The most popular scoring convention in goal setting is a laundering operation. It was not designed as one. The best laundering operations never are.

The convention says: score each Key Result from 0.0 to 1.0, and treat 0.7 as success. On paper this protects ambition. Nobody gets punished for reaching, so everybody reaches.

In practice, teams learn the real game within two quarters. When a number carries consequences, people do not get braver. They get better at arithmetic. So you write targets you privately expect to miss by about thirty percent, you land where you always knew you would, and you collect the label "ambitious" on the way through. The sandbag goes in dirty. The arithmetic washes it. It comes out reading as stretch. Nobody lied. Everybody negotiated.

There is a tell, and you can check it this afternoon: pull your last quarter's scores. If they cluster between 0.6 and 0.8, you are not looking at performance. You are looking at negotiation, laundered clean.

One more receipt, from issue one. When the machine cloned our methodology, it threw away our grading system and installed 0.7 in its place. In our voice. Confidently. The convention is so deep in the average that even our own copy defaulted to it.

What we run instead takes two instruments, because one number was never enough. Targets are calibrated to be hit: one hundred percent means what it says, and where a target deliberately stretches it wears a label, aspirational, so nobody pays for the reach. Then the quarter is graded, Excellent to Bad, by judgment with the evidence on the table. Fifty percent progress can grade Excellent if the learning was worth the quarter. A hundred percent can grade Poor if the target was a sandbag all along.

This fortnight we wired the grades to money, carefully, and the word carefully is doing real work. Team OKRs never touch bonuses. Not ever. Cash on a team's goals kills the ambition and the candour the whole system runs on, so at team level the grades stay what they should be: learning, on the record. The only grades that reach money belong to the SLT's few strategic OKRs, read together at company level as a gate. Those retrospectives are moderated by named humans with the rationale written down, and the mapping from grade to payout is published at plan launch. In the changelog it reads: the gate reads judgment, not arithmetic, and it only reads the top. Grades decide. Percentages inform. Teams stay free to stretch.

Take last quarter's scores into Monday's meeting and ask one question: which of these targets did anyone believe on the day they were written?

Grade, don't score.

See you in a fortnight,
Matt

P.S. The free AI OKR Coach kit flags sandbagged targets on sight, along with the other habits the average teaches. No charge, nobody will call you.