He had the better year. She had the better record.
Two candidates, one band promotion, near-identical delivery. The panel never saw the work itself, only two accounts of it.
Both were senior analysts in the same function. Both had delivered. If you had watched the year unfold from the next desk, you would have said he had the stronger twelve months: he ran the harder migration, he took the weekend calls, he inherited the worst of the legacy system and shipped anyway.
Neither of those years went into the room. What went into the room was six pages each.
What his six pages said
His case was written the week it was due, from memory and a scan of his sent mail. It read well. It was also almost entirely adjectival: he had "led a complex migration," "partnered closely with stakeholders," "drove significant efficiencies."
Every one of those statements was true. None of them could be checked. When a panel member asked what "significant" meant, he gave a range, then corrected himself, then said he could find out.
"I don't distrust the claim. I just can't defend it upstairs if I can't cite it."
What her six pages said
She had kept a log for two years. Thirty-one entries that year, most of them four lines long, each written within a week of the event. Her case was not better written. It was better sourced.
Where he wrote "drove significant efficiencies," she wrote that a scored intake checklist replaced a manual review in March, that median intake time fell from eleven days to three by June, and that the ops dashboard and two named stakeholders could confirm it. The panel checked one of the three claims, found it accurate, and stopped checking.
That is the whole mechanism. Verifiable specificity buys trust, and trust gets extended to the claims nobody bothers to verify.
The part that isn't about writing
Nine months before the review, she had logged the intake project as it happened, including the two weeks it looked like a failure. He had done comparable work and kept no trace of it. The asymmetry wasn't effort or ability or confidence. It was a Friday habit.
He got the band the following cycle, incidentally. He started a log in the second week of January.
What to do with this
Take one project you finished more than three months ago and try to write four lines about it: what you decided, what changed, what number moved, who saw it. If that takes more than five minutes, the problem isn't your memory. It's that nobody wrote it down at the time, and the only person who can is you.