Hiring.

The framework

1. 01

   Ask for the day, not the adjective.

   The whole method lives in one question, asked over and over: tell me about a specific time this job went well, and a specific time it went badly. What was happening beforehand, what did the person actually do, and what followed? Push until you've got a scene with a date attached. When somebody offers you 'she was proactive', that's a summary of an incident they haven't told you yet, so ask what she did that made them reach for the word. The story is the data; the adjective is what's left after the data evaporated.

2. 02

   Twenty stories, four directions.

   One person's account of a job is one person's theory of it. Collect roughly twenty incidents from four directions — whoever has done the role, the manager it reports to, the people who work beside it every week, and the one everybody skips, the customer or internal client on the receiving end. Ten going right, ten going wrong, about five from each direction. Flanagan's own studies ran to hundreds; twenty is the small-company floor, and you'll know you've reached it when a new story stops introducing anything you haven't already heard. That arrives sooner than you fear and always later than the impatient version of you wanted.

3. 03

   What did the difference actually turn on?

   Lay the stories out and sort them by what separated the good day from the bad one. You're looking for the handful of behaviours that keep recurring on both sides of the pile — the person who rang the client before the client rang them, the one who let a small discrepancy in the numbers go past. Two or three clusters will do most of the work; a list of eleven means you're sorting by topic and not by consequence. Those clusters, not the org chart and not last year's advert, are what this role requires.

4. 04

   Test the wish list against the file, then hand what survives forward.

   Now run the usual requirements past the incidents: the degree, the years of experience, the industry background, the named software. If one of them never appears in a story as the thing that made the difference, it's a preference you've been enforcing as a rule, and it's been quietly narrowing your field for years. Cut it here, once, and don't relitigate it later in the process. What's left — the clusters plus whatever the evidence genuinely supports — is the competency list tomorrow's scorecard gets built against, and the file itself becomes the questions you ask and the standard you mark answers by.