---
title: Classic vs Rolling vs Cumulative Retention Explained
description: Classic, rolling and cumulative retention count different users on the same day. See the definitions, a ten-user worked example and when to use each one.
url: https://retentioncohorts.com/learn/classic-vs-rolling-vs-cumulative/
---

[Home](https://retentioncohorts.com/) / [Guides](https://retentioncohorts.com/learn/) / Classic vs rolling vs cumulative retention

# Classic vs rolling vs cumulative retention

Classic, rolling and cumulative retention count different users on the same day. See the definitions, a ten-user worked example and when to use each one.

**Short answer**

Classic retention counts users active on exactly Day N, rolling counts users active on Day N or any later day, and cumulative counts users active at least once from Day 1 to Day N. All three use the same install cohort, so they give different numbers for the same app on the same day.

## What are classic, rolling and cumulative retention?

Classic, rolling and cumulative retention are three ways of counting how many users from one install cohort came back, and they differ only in which days count as a return for Day N.

Every version starts from the same place: a cohort of users who installed on the same day. The install day is Day 0, and activity on Day 0 is never a return. Day N retention is the users counted as retained on Day N divided by the installs in the cohort.

**Classic retention:** the share of the cohort active on exactly Day N. Also called [N-day retention](https://linkrunner.io/glossary/what-is-n-day-retention?utm_source=retentioncohorts&utm_medium=referral&utm_campaign=learn-classic-vs-rolling-vs-cumulative&utm_content=inline) or exact-day retention.

**Rolling retention:** the share of the cohort active on Day N or any later day observed so far. Also called unbounded or "on or after" retention.

**Cumulative retention:** the share of the cohort active at least once from Day 1 to Day N.

| Definition | Counts a user on Day N if they were active... | Bounded? |
|---|---|---|
| Classic | on exactly Day N | Yes: fixed once Day N has passed |
| Rolling | on Day N or any later observed day | No: rises as late returns arrive |
| Cumulative | at least once from Day 1 to Day N | Yes: fixed once Day N has passed |

The [retention types page](https://retentioncohorts.com/types/) gives each definition its own section with formulas and edge cases, and the [cohort explorer](https://retentioncohorts.com/) lets you switch one table between all three.

## A worked example with ten users

The [ten-users view on the home page](https://retentioncohorts.com/#user-dots) uses a made-up group of ten users, observed for 14 days. Some open the app every day, some once, one never returns.

|  | Classic | Rolling | Cumulative |
|---|---|---|---|
| Day 1 | 5 of 10 (50%) | 9 of 10 (90%) | 5 of 10 (50%) |
| Day 7 | 3 of 10 (30%) | 6 of 10 (60%) | 8 of 10 (80%) |

Why the Day 7 numbers differ:

- **Classic** only sees the three users who opened the app on Day 7 itself.
- **Rolling** also counts users who skipped Day 7 but came back on Day 8, 12 or 14.
- **Cumulative** counts everyone who came back at least once in the first week, even if they never returned after Day 2.

Same users, same app, three honest answers to three different questions.

## Why is my rolling retention higher than classic?

Rolling retention is always at least as high as classic because it counts every user classic counts, plus anyone who missed Day N but returned later. The gap is widest for apps people use on irregular days.

Cumulative is also always at least as high as classic. Rolling and cumulative, however, can be in either order: in the ten-user example rolling is higher on Day 1 and cumulative is higher on Day 7.

## Bounded vs unbounded: why rolling keeps moving

Classic and cumulative are bounded. Once Day N has passed for a cohort, its value is fixed and will never change.

Rolling is unbounded. It looks forward from Day N to the last day your data covers, so it keeps rising as late returns arrive. That has two practical effects:

- A rolling number you exported last week is already out of date.
- An older cohort has had more days for users to return than a young one, so its rolling value looks better even if the users behaved the same.

There is a longer explanation of this in [why rolling retention keeps rising](https://linkrunner.io/blog/why-rolling-retention-keeps-rising?utm_source=retentioncohorts&utm_medium=referral&utm_campaign=learn-classic-vs-rolling-vs-cumulative&utm_content=inline).

## When to use which definition

### Classic

- Comparing against published benchmarks, which usually use classic (exact-day) retention.
- Measuring the effect of a product change, push or release on a specific day.
- Comparing cohorts with each other, because the value is fixed once the day passes.

### Rolling

- Answering "are these users still around at all?"
- Apps used on irregular days, such as travel, shopping or finance apps, where exact-day retention looks harsh.
- Always refresh it before you read it, and never compare a young cohort with an old one on rolling.

### Cumulative

- Activation and onboarding questions: did we bring them back at least once by Day 3 or Day 7?
- Judging the first week of a new user journey, where any return counts as success.
- It never falls as N grows, so do not read a rising cumulative curve as improving engagement.

A longer comparison is in [this guide to cumulative, classic and rolling retention](https://linkrunner.io/blog/cumulative-classic-rolling-retention?utm_source=retentioncohorts&utm_medium=referral&utm_campaign=learn-classic-vs-rolling-vs-cumulative&utm_content=inline).

## The mislabelled column trap

The most common mistake is not choosing the wrong definition but not knowing which one a column uses. A dashboard labelled "D7 retention" could be any of the three.

- If a number looks too good, check whether it is rolling or cumulative before celebrating.
- If two tools disagree, check the definitions before you check the data.
- If you calculate it yourself, the [SQL and spreadsheet formulas](https://retentioncohorts.com/compute/) for all three are on one page.

Attribution tools and analytics tools differ here, so read the docs for whichever tool you use. If you are new to attribution tools in general, [this explainer on what an MMP is](https://mmpcompares.com/blog/what-is-an-mmp/) covers the basics.

**Note:** Example numbers come from a made-up app to show the method, not a benchmark.

See this on your own app

Linkrunner reports classic, rolling and cumulative retention as separate columns, so the question of which definition a number uses has a clear answer for your own app.

- [Cumulative vs Classic vs Rolling Retention](https://linkrunner.io/blog/cumulative-classic-rolling-retention?utm_source=retentioncohorts&utm_medium=referral&utm_campaign=learn-classic-vs-rolling-vs-cumulative&utm_content=panel)
- [Create a free account (25k installs free)](https://dashboard.linkrunner.io/sign-in?utm_source=retentioncohorts&utm_medium=referral&utm_campaign=learn-classic-vs-rolling-vs-cumulative&utm_content=panel)

## Questions

**Is classic retention the same as Day N retention?**

Yes. Classic, N-day and exact-day retention all mean the share of the cohort active on exactly Day N. It is the definition most published benchmarks use.

**Can rolling retention be lower than cumulative?**

Yes. Rolling and cumulative answer different questions, so either can be higher. In the ten-user sample, rolling is higher on Day 1 and cumulative is higher on Day 7.

**Why did my rolling retention change for an old cohort?**

Rolling counts returns on any day after Day N up to the last day observed. When a user from an old cohort comes back today, every rolling value up to that day rises, so the number keeps moving.

**Does Day 0 count as retention?**

No. Day 0 is the install day, and activity on that day is not a return. All three definitions start counting from Day 1.

**Which definition should I report to leadership?**

Report classic for trend lines and benchmark comparisons, because it is fixed once the day passes. Add cumulative if the question is about activation, and label every column with its definition.

**What is the difference between classic, rolling and cumulative retention?**

Classic counts users active on exactly Day N. Rolling counts users active on Day N or any later day, and cumulative counts users active at least once from Day 1 to Day N. They use the same raw activity but answer different questions.

## More guides

- [01 **How to read a retention curve** A retention curve plots one cohort's retention against Day N. Read the early drop, the flattening point and smile curves, and why shape depends on definition.](https://retentioncohorts.com/learn/read-a-retention-curve/)
- [02 **Why paid D1 retention is lower than organic** Paid D1 retention below organic is partly normal, as organic users arrive with intent. See the causes, from targeting to measurement errors, and how to check.](https://retentioncohorts.com/learn/paid-d1-lower-than-organic/)
- [03 **Retention by channel and campaign** Blended retention hides what each channel does. Learn to cut a cohort table by channel, campaign and creative, keep naming clean and spot a channel mix shift.](https://retentioncohorts.com/learn/retention-by-channel/)
- [04 **Choosing the cohort start event: install or first purchase** How to define cohorts by choosing a start event: install, first open, sign-up or first purchase, what each answers, survivorship bias, day boundaries, channels.](https://retentioncohorts.com/learn/cohort-start-event/)
- [05 **Retention for weekly-use vs daily-use apps** Why classic Day N retention undercounts apps people use weekly, and how cumulative windows, weekly cohorts and DAU/MAU give a fairer read of weekly-use apps.](https://retentioncohorts.com/learn/weekly-vs-daily-use-apps/)
- [06 **Turning retention into decisions: LTV, payback and budget** How to turn app retention into decisions: which definition feeds LTV and forecasts, a simple LTV from a retention curve, cost per retained user and payback.](https://retentioncohorts.com/learn/retention-to-decisions/)