The mock meeting and the one-page brief

False beliefs first. Any number above zero means a leading question slipped through, and on this bank each one is a modelling decision you will get wrong in week six. "So you want to forecast sales, right?" records the costliest of the five: the target is a stock-out, not sales, and last year's vendor already built the sales forecast nobody used. Find the leading question in the transcript, write the open version on your sheet, and carry that version into the real meeting.

Doors opened. Six is the only answer that lets you write a brief without guessing. A shut door is a block of the brief you cannot fill. In the 200 shuffles on this bank, the doors most often left shut are constraints and decision, and decision is the one a scientist cannot afford: it holds the Thursday van, and a prediction that arrives on Friday is useless.

Net value against the ladder. The Six Doors in printed order score 46 blind; tuned for a scientist, outcome first and widest, 55. A run in that region means the sheet works. A run near the shuffle's 27.5 means the questions are good and the order is not: usually the outcome door was opened late, or the technical opener crept in early and spent trust the rest of the meeting needed. The technical opener, labels and class imbalance first, scores 10, below the efficient interview's 15.

The 20-minute counter. The tuned order puts 24 points in your notes by minute 20; the printed order, 20. If she is called away halfway, that is the meeting you actually had. The efficient interview sits at -5 at minute 20, because its five wrong beliefs are recorded in the first ten minutes and its facts come later.

The bench scores what you asked and what came back. It does not hear your tone, see whether you looked at her or at your laptop, or notice the moment she leaned forward because a question was about her actual Thursday. Those matter, and no widget measures them honestly. The run is a rehearsal for the order and the kinds, not a verdict on you as an interviewer.

Even the tuned order leaves gaps, and they matter for the brief. It asks eleven open questions in 45 minutes and stops before the lead-time question, the "when do you next have to show something upward?" question and the question about who holds the POS login. So its notes have no five and nine days of supplier lead time, no six-week board date, and nothing about Kenneth. The brief must not write those as facts. They go in the covering note as questions, which is where session 5 said they would go.

1 · What you told me, in your words. Two or three sentences she said, quoted. Not your summary, and not your restatement of the target. "We keep getting caught" and "a patient turns up and the shelf is empty" are hers; "predict stock-out events at outlet-product grain" is you tapping the tune. The base rate goes here too, in the form she gave it: roughly 30 a month across the 38 outlets, and she knows because each one becomes a complaint call. If you misheard, this is the block she corrects, and it costs one email instead of six weeks.

2 · The decision this will serve. The action, the person, the deadline, from door 2. One line: the area manager moves stock between outlets on Thursday's van run, per outlet per product, once a week. If you cannot write it, you did not open the door, and the brief should say "I still need to understand who acts on this and by when" rather than invent a daily forecast nobody asked for.

3 · What has been recorded, and for how long. Session 2's ladder, as five lines, each marked recorded, typed or not recorded, with the baseline written in: today the pharmacist looks at the shelf and WhatsApps the area manager, and that catches most of them. End with one question: "which of these is wrong?"

4 · What I will build first, and what I will not. The pilot from session 5 in one sentence: a daily stock-out log in four outlets, the Johor pharmacist first. Then one sentence naming what you are deliberately not doing yet: no model, until eight weeks of logged stock-outs exist and a plain rule has a record to beat. Saying "no model yet" out loud, to a person who asked for a prediction and a board that wants the word AI on a slide, is the most useful line in the brief. It is also the line that most needs her agreement, so ask for it by name.

5 · How we will both know. Her success test, in her words: the complaint calls about empty shelves go down, and she counts them herself. Then the guard rail, written as your reading for her to correct: an over-order costs expiry write-off, which head office measures her on, so the calls must not come down by filling every shelf. Both costs of a wrong guess, in one block. Not an accuracy figure.

6 · Who does what next week. Names, a day, a time, one task each. Fitzpatrick's test for whether a conversation went anywhere, as reported, is whether the other side committed something: time, a name, an introduction. The Johor pharmacist already did, before you asked; this block is where that offer, and Priya's two hours a week, are written down so they cannot quietly lapse.

Within a day she still remembers what she meant, so a correction is cheap and specific. After a week she remembers that you came by, and a brief arriving then reads as a proposal to be evaluated rather than a record to be checked. No study is quoted for the 24 hours; it is a working norm, and the reason is the one above.

One page because she will read it on her phone. Her area managers read everything on their phones, in the van, and so does she between outlets. Two pages is a document; a document goes into a folder to be read later; later is after the board. If a block needs more than three lines, the detail goes in an attachment she can ignore.

A word from the jargon column of session 4: labels, class imbalance, precision, recall, a holdout. An accuracy figure, promised or estimated, for a model that has nothing to learn from yet. A number you have not seen stated as if you had, including the supplier lead times and the board date if your run never reached them. A timeline longer than the date she gave you. And a fact she told you in confidence, most importantly the one about Kenneth leaving in January, if trust ever ran high enough for her to say it: it shapes block 4, because the first step must not depend on a login that disappears, but it goes into your plan and not into a page she may forward to a team she has not told. The brief is hers to forward. Write it as if it will be.