留存掉了，怎么查
次日留存从 42% 掉到 35%，你第一步查什么？
Copy the prompt, then pick a model. It plays the interviewer and asks you through it scene by scene instead of handing you the answer.
On a phone with the app installed, the system opens the app directly. Links cannot open external apps inside WeChat — open it in a browser first.
The copied text ends with the answer key the model grades you against. Don't read it yourself first — working it out on your own is the point.
你一上来说「是不是因为改版了」，它会打断你：你还没看数据就有结论了？
这一份的中心是一个很多人会跳过的动作：把新用户和老用户分开看。新用户的留存天然更低，所以新用户一变多，整体留存就会「下降」——而每一群人的留存其实一点没变。增长期几乎必然踩到这个。
第四幕会让你只看形状不看数值 —— 留存曲线的三种形状意义完全不同，而其中一种指向这批题里最值钱的那个问题。
What this one tests
This prompt in the repo (the answer key is at the end of that file — you'll see it)
Have you done this one?
Sign in to mark this one as practised — next time you come back you'll see when you did it.
Notes from people who did this one
How to use this
There is no model answer here — but there are definite wrong answers. Only three things make a case hold up, and every prompt's answer key carries a "common wrong answers" section that is more useful than the checklist itself.
The three things that make it hold up
- You can say whose problem this is and what it is — **with evidence** (not "I think users would want it")
- You can say what success looks like — **and what that number is today** (no baseline, no acceptance criteria)
- You can name the cost and what you gave up — **the hardest one to fake**, which is why every prompt hunts for it
How to start
- Pick a question, open it, copy the prompt.
- Paste it into a brand-new conversation — not one where you already discussed it.
- Say nothing. Wait for its first question.
Four commands you can call out at any time
跳过move to the next scene now提示give a direction, not the answer直说吧spell this point out in full, then keep pressing on the trade-offs复盘jump to the wrap-up and print the review card
Three traps
- Don't rush to a solution. More than half the lost points here come from designing before asking — especially describing features before defining what success means.
- Expect to be pushed for numbers. "It'll perform better", "users will like it" get cut off: which metric, what is it today, and when do you look?
- When it starts praising you, point it back to rule five — feeling validated crowds out noticing what you missed.
Start with the north star metric; it underpins everything else. Then pick by what's on your plate: a backlog you can't order → prioritisation; nobody using it or retention sliding → retention and onboarding; need to price something → pricing. Save the experiment review for later, and keep redoing the one about turning a one-line request into a spec — it's the closest to daily work.
Sign in to write a note.