> ## Content Index
> Fetch the complete content index at: https://launch.ghostcms.templates.codememory.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# How to read a retention curve
- URL: https://launch.ghostcms.templates.codememory.com/how-to-read-a-retention-curve/
- Published: 2026-07-02T14:00:00.000Z
- Updated: 2026-07-02T14:00:00.000Z
- Description: The cliff, the slope and the flat line: what each shape means, and what to work on next.
- Author: Maya Chen
- Tags: Guides, #Import 2026-10-02 15:41

Retention is the share of people who come back after their first visit. It is the most honest number in product analytics, because it is hard to fake. You can buy signups, but you cannot buy people who return every week because the product helps them. This guide explains how to read a retention curve, and what to do with it.

![A retention curve: steep at first, then, if you are lucky, flat.](https://launch.ghostcms.templates.codememory.com/content/images/2026/10/ret-chart.jpg)

A retention curve: steep at first, then, if you are lucky, flat.

## What the curve shows

A retention curve starts at 100% on day zero: everyone who signed up. Then it falls. Some people try the product once and never come back. Others return for a week and drift away. A few stay.

The shape matters more than any single point. There are three common shapes:

1. **The cliff.** The curve falls to near zero within a few weeks. People are not finding value, or not finding it again.
2. **The slope.** The curve keeps falling slowly, week after week. People like the product but have no reason to make it a habit.
3. **The flat line.** After the early drop, the curve levels out. A stable group of people keeps coming back. This is what you want.

## Choose the right "return"

The most common mistake is counting the wrong thing as a return. Opening the app because of a notification is not the same as using it. Choose an action that means value: a report viewed, a message sent, an order placed. In Tally, you choose this as the return event.

The second choice is the time period. A product used daily, like a messaging app, needs daily retention. A product used for monthly invoicing needs monthly retention. Using the wrong period makes a good product look bad, or a bad one look good.

> Measure people against the rhythm of the problem you solve, not against the rhythm you wish they had.  
>  
> **Maya Chen**

## Compare cohorts, not averages

A cohort is a group of people who started in the same week or month. Comparing cohorts shows whether your product is getting better. If the people who joined in March stay longer than those who joined in January, something you changed is working.

| Cohort   | Week 1 | Week 4 | Week 8 |
| -------- | ------ | ------ | ------ |
| January  | 42%    | 21%    | 15%    |
| February | 45%    | 24%    | 18%    |
| March    | 51%    | 31%    | 26%    |

An average over all users hides this. Old, loyal users make the average look healthy while new users leave.

![Habits form in the first weeks. That is where retention is won or lost.](https://launch.ghostcms.templates.codememory.com/content/images/2026/10/ret-phone.jpg)

Habits form in the first weeks. That is where retention is won or lost.

## What to do with it

If your curve is a cliff, work on the first visit: what does a new person need to see or do to get value? If it is a slope, work on the reason to return: reminders, saved work, something that changes each week. If it is flat, work on getting more people onto it, because now growth will actually stick.

💡

Look at retention once a month, not once a day. Daily numbers are noise; cohorts need time to tell their story.

#### What is a good retention number?

It depends on the product. Compare against your own earlier cohorts first, not against other companies.

#### Should I count people who never activated?

Show both. Retention of activated people tells you about the product; retention of everyone tells you about your onboarding.