Hiring.

The framework

1. 01

   Start from the exits, not the advert.

   The honest preview isn't invented in a meeting. It's collected from people who have already left. Ask the last two or three who held this role, or one close to it, what they'd have wanted to know on day one — and ask them once they're safely elsewhere, where the answer costs them nothing. You'll get things no advert has ever carried: the client who calls at weekends, the six weeks a year that eat everything, the support that was promised and never quite arrived. That is your preview. You didn't write it; you just stopped hiding it.

2. 02

   Say it live, and let them see the real thing.

   Phillips found the medium mattered, and not in the direction a production budget would suggest: spoken and written previews both did more to cut turnover than the videotaped kind, and the video is the expensive one to make. Whatever advantage it held showed up on how people performed, not on whether they stayed. So the sentence in the advert is doing real work, and so is the person who says it out loud on the first call — run both. Then go past what either of them can carry. Let the candidate spend an hour with whoever does the job now, unsupervised and free to say anything at all. Let them see the backlog, the ticket volume, the room. What you're testing is whether reality changes their answer, and reality is hard to stage.

3. 03

   The sentence you're most afraid to publish.

   There is always one — the fact you believe would empty your pipeline the moment it appeared in writing. It might be eleven months of cash runway, or a founder who rewrites finished work at midnight, or a role that is sixty per cent administration for its first year. Whatever it is, that's the sentence doing the most work in your whole preview, because it's the one that sorts hardest. Publish it plainly, without apology or padding, and watch who stays. The people who read it and still want the job are exactly the people you were looking for.

4. 04

   Ninety days is the number that grades the preview.

   You'll be told an honest preview thins the pipeline, and Wanous's self-selection argument does predict exactly that — but Phillips's data didn't find it, so don't grade the thing on a theory. Grade it on the number it demonstrably moves. Count how many of your hires are still there at ninety days, and of those who left, how many cited something you already knew in week zero. That second figure is the one that should fall to nothing. Each of those exits costs you a search, a salary, a manager's year and whatever the rest of the team quietly concluded about how long people last here.