Retention falls

The retention drop needs to be localized first, not explained. Check the definition and completeness of the data, start date, segments, versions and channels; then match the change with releases, service quality and audience composition, retaining several competing hypotheses.

How to know if retention is really falling

Check the data.

First, make sure that the retention drop is not a data bug or an artifact of changes in analytics. Compare definitions: whether you have the same window (for example, day 1, day 7 or month to month), whether cohorts are correct, whether the application has updated, or whether the event tracker has changed.

**Example: Projects often experience a sharp drop due to incorrectly counted returns after an application update or errors in the analytics config.

Find out what retention is falling.

Retention is counted by different windows: day 1, day 7, day 30, monthly retention, rolling retention.

**Antipattern: running hypotheses to the touch when it is unclear, retention in which cohort and after how many days it has subsided.

Quick Diagnosis: What to Watch First

See the slices: platform, segment, channel

Retention rarely falls across all segments at once. Quickly decompose the retenshen in the section - platform (iOS, Android, Web), attraction channel (how new users got), segment (new vs paying).

**Example: Often the problem is with only one source of traffic – for example, a new channel leads to low-quality users whose retention is below normal.

Compare changes in the product core

Check if there were changes to the product or the login funnel: new onboarding, subscription changes, content closures, it became harder to sign up, compare engagement rates for key actions before and after the change.

The most common causes of retention

The problem with first-hand experience

The user does not understand why to return, did not see quick benefits, did not understand the interface, did not get the desired result.

Onboarding failure

Complex registration, a lot of mandatory data, bugs, unnecessary steps.

The mechanics of engagement are outdated or have lost relevance

The replay is boring, there is no personalization, no return triggers (push, emails don’t work).

Example: The expense app stopped reminding you to make a spending call.Push notifications were turned off at backend, retention broke in less than a week.

Attracting non-convertible users

The advertising campaign or the traffic channel has changed, and there are users who don’t need the product, and the overall retention is falling because of the erosion of the core of the audience.

Quick Actions: What to Do in the First Week

1. Detail the fall in slices

Build retention by generation, sign-up time, source, platform, account age, quickly understand where the problem is most acute.

2. Check the user funnel

Look at the key steps: registration, first login, first target action. Retention often sags if something breaks or changes here.

** Case: After simplification of on-boarding, the number of registrations increased, but retention day 1 and day 7 fell - new users did not know what to do next.

3.Contact users

Gather fresh feedback: post-action polls, letters to loyalists, calls to the departed. It’s important to understand what pushed you away – you better hear it live.

4.Compare product changes with the moment of the fall

Check out any releases, bugs, and updates related to the core of value.

5. Check for technical failures

Whether server requests are falling, whether flies are being sent on time, whether there is an increase in errors in the application, sometimes the retention drop is a bug, not a product problem.

What to rely on in making decisions

Metrics and sources for benchmarks

Retention day 1, day 7, day 30, rolling retention and cohorts by month answer different questions. Choose a window by natural frequency of product value and compare comparable cohorts; an average benchmark of another category will more likely confuse diagnosis.

Recommendations and cases can be found in:

• Amplitude Guide to Retention
• Mixpanel Retention Benchmarks

Important: look not only at the total number, but at the dynamics of cohorts.

Expert principles

Anti-patterns: What NOT to do when retention falls

• Run email and fluff without analyzing the cause
• Scaling new hypotheses to all without breaking down segments
• Ignore bugs and technical problems
• Blindly copy someone else’s experience without considering the characteristics of your audience

Example: The team runs bonuses and kicks to the entire base when retention drops, although the problem is in one Android segment without access to an important feature.

FAQ

What is the key retention? The most important slice is the one that reflects the value of the root scenario: day 1, day 7, or monthly. Focus on your product and how often you use it.

Where does retention usually fall from: bug or product? Half of the time, it’s due to bugs or technical failures, and the other half is product changes, loss of value at the start, or poor quality traffic.

What are the first steps to reduce retention? Check the data, break the metric into platforms, channels and segments, analyze the funnel, find out what the departing users say.

What to look for in analytics, except retention? Analyze cohorts, conversions to key actions, number of bugs, depth of sessions, feedback of users.

Can I quickly fix retention push notifications? If the problem isn’t the mechanics of the returns, but the mistakes or loss of value, no mailings will save Push is doping, not a cure.

Where to find good retainers and good retainers? On Amplitude Blog and Mixpanel SaaS Benchmarks