Building the Story Bank

The bank is a private document — not part of this map, and not something to publish. Its purpose is to make recall cheap under stress, because the failure in the room is almost never "I have no good stories", it's "I couldn't retrieve the right one in four seconds and told a weak one instead."

Sourcing. Go through the last two or three years by artefact rather than by memory: PR history, design docs, incident channels, performance review notes, the ticket you were dreading. Memory over-indexes on recent work and under-indexes on the hard thing you did eighteen months ago. If you keep a brag document, this is what it's for; if you don't, start one now for the next search.

Coverage. Aim for at least one story each for: a project you owned end to end, a disagreement with someone senior, something that failed and what you changed, work in real ambiguity, mentoring or levelling someone else up, a hard technical debugging or design win, a time you pushed back on scope, and — for these companies specifically — something you built with or around models. If two slots are filled by the same project, that's a gap, not efficiency.

Format. Half a page each: one line of situation, your scope, three or four decision bullets, the result with a number where one exists, and what you'd change. Then a tag line — ownership, judgement, conflict — because retrieval in the room happens by signal, not by project name.

Rehearsal. Say them out loud, ideally recorded, once each. Not to memorise — memorised stories sound memorised and collapse under follow-up — but to find the places where you ramble, and to get the situation down to two sentences. Then rehearse the index: given "tell me about a conflict", which two stories are candidates.

Keep it current. The best material is from the job you're in now, and it's the material you'll have forgotten first.

Behavioural questions are drawn from a small pool with a large number of surface phrasings. Preparing per-question is unbounded work; preparing eight stories with good tags is a weekend, and it's the highest-return preparation in the whole loop. It also feeds the CV, the LinkedIn summary, and the deep-dive round from the same source.

• Eight to ten stories, half a page each, is enough to cover a full loop.
• Source them from PRs, docs, and incident channels — memory skews recent and flattering.
• Cover ownership, conflict, failure, ambiguity, mentoring, a technical win, a scope push-back, and something AI-adjacent.
• Two stories drawn from the same project is a coverage gap, not reuse.
• Tag each story by the signal it demonstrates; retrieval in the room is by signal, not by project.
• Rehearse out loud once to tighten the setup, never to memorise wording.
• The same document feeds your CV, your positioning line, and the deep-dive round.
• Keep a running brag doc so the next search starts with material instead of archaeology.