# Find the activation metric that predicts customer retention

A method to identify the early actions that separate retained users from churned ones, and how to drive them.

- Topic: Data & Analysis
- Framework: Few-Shot
- Level: Standard
- Tags: HR & Hiring, Fill-in Template
- Source: https://www.thinkwithprompt.com/prompts/find-the-activation-metric-that-predicts-customer-retention

## Prompt

```text
Help me find the activation moment for my product: the early actions that predict a new user or customer sticks.

**Product and core value:** [WHAT it does and for whom]
**What "retained" means for us:** [e.g. "still active in week 8", "renewed after year 1"]
**Events I track in the first days/weeks:** [LIST: e.g. created project, invited teammate, connected integration, sessions in week 1]
**Data available:** [PASTE a summary by cohort or describe what I can export]
**Current activation definition, if any:** [DEFINITION and rate]

Steps:
1. List candidate activation events and combinations (e.g. "invited 2+ teammates in first 7 days").
2. Explain how to compare retention for users who did vs. didn't do each action, by cohort, and how to pick a threshold (the point where retention gains flatten).
3. If I've pasted data, run the comparison and show it.
4. Warn about correlation vs. causation: power users do everything. Suggest an experiment that pushes some users to the action to test whether it causes retention.
5. Recommend the activation metric, its target time window, and a target rate.

Then give me growth ideas to increase it: onboarding steps to remove, emails or in-app prompts to add, and sales/success touchpoints, each tied to the metric.

Return a short memo plus a table: Candidate action | Retention if done | Retention if not done | Gap | Notes. Don't fabricate figures; mark missing data [NEED].
```

## What to fill in

- WHAT it does and for whom
- e.g. "still active in week 8", "renewed after year 1"
- PASTE a summary by cohort or describe what I can export
- DEFINITION and rate
- NEED

Free to copy and use. From ThinkWithPrompt (https://www.thinkwithprompt.com/), a library of 2,663 curated AI prompts.