Retention shows you how many users return during subsequent periods, according to PostHog.
A retention chart labels return periods in the units the publisher names, and each publisher states its own rule.
Retention charts track user return activity across defined time intervals. Each analytics platform defines these intervals using specific units and rules. PostHog names its units as hours, days, weeks, and months. Mixpanel uses day buckets, such as day 5 or ranges like 15 to 30 days. Amplitude compares a starting event date to a return event date. Google Analytics provides cohort examples using weekly figures.
You must identify the specific unit a chart uses before interpreting the data. A chart labeled in days measures different intervals than one labeled in weeks. The rule for counting a return varies by publisher. PostHog states that the first period always shows 100%. Mixpanel describes day 5 retention as the percentage of users who came back and did B exactly on the fifth day after the A event. Amplitude defines retention based on the time between the starting and return events. Google’s documentation uses a cohort example with 60% and 25% weekly figures.
Key takeaways for retention period labels before a health review
- PostHog states that the first period is always 100% because all users in that cohort must perform the event to be included according to PostHog.
- Mixpanel defines day 5 retention as the percentage of users who came back and did B exactly on the fifth day after they did the A event according to Mixpanel.
A filled reference table of retention period labels
The table below lists the four publishers and their stated rules for labeling return periods.
| Publisher | Period labels or units | Stated rule or definition |
|---|---|---|
| PostHog | Hours, days, weeks, months | Define the retention period by any number of hours, days, weeks, or months, according to PostHog. |
| Mixpanel | Day 5 | Day 5 retention is the percentage of users who “came back and did B” exactly on the fifth day after they did the A event, according to Mixpanel. |
| Amplitude | Starting event and return event | Amplitude computes retention data by comparing the date of that starting event to the date of the return event you specified, according to Amplitude. |
| Cohort grouping | Cohort reports analyze user retention over time by grouping users with a shared characteristic, according to Google. |
Use this table to verify the labels on your current chart before the health review.
PostHog's hours, days, weeks, and months
PostHog defines the retention period by any number of hours, days, weeks, or months, according to PostHog. These four units constitute the specific labels you must identify on the chart before the health review. The documentation states that the first period is always 100% because all users in that cohort must perform the event to be included, according to PostHog. This rule applies to the first period, since all users in that cohort must perform the event to be included.
When reading subsequent periods, PostHog provides two distinct calculation examples to clarify how percentages are derived. For example, if Day 3 shows 25% for a cohort, it means that 25% of the users who performed the start event returned on Day 3, according to PostHog. Alternatively, with the previous-period option, if Day 3 shows 40%, it means that 40% of the number of users who returned on Day 2 returned on Day 3, according to PostHog. This second metric calculates retention based on the users active in the immediately preceding period rather than the initial cohort.
Understanding the difference between these two calculation methods is critical for accurate interpretation. The first method is the share of users who performed the start event and returned on that day, while the second is the share of users who returned on Day 2 and then returned on Day 3. Both examples use Day 3 to illustrate the mechanics, but the underlying logic changes depending on which option is selected in the chart configuration. The 25% figure and the 40% figure are examples provided by PostHog to demonstrate these definitions; they are not measured results for your specific account or customer base.
Before proceeding to the health review, verify which calculation method your current chart uses. Check which option is selected before the health review: the return on that day or the previous-period option. Confirming this distinction prevents misreading the trend line during the review. The four units—hours, days, weeks, and months—determine the granularity of these periods. Selecting the appropriate unit ensures the chart aligns with the typical usage cycle of the specific product being reviewed.
Mixpanel day buckets, including day 5 and the 15-to-30-day range
Mixpanel defines retention using specific day counts to measure user return patterns. According to Mixpanel, day 5 retention is the percentage of users who “came back and did B” exactly on the fifth day after they did the A event. This rule ties the return action to a precise calendar day rather than a broader window.
The source also illustrates how to frame questions for mid-term engagement. According to Mixpanel, the tool answers an important mobile gaming question: How often do my users come back and play a game between 15 and 30 days after signing up to represent mid-term players? This example shows how a range of days can be used to identify a specific segment of users who return after the initial sign-up period. It serves as a model for defining mid-term retention windows in your own analysis.
You can also adjust how daily data is presented. According to Mixpanel, you can group daily buckets to understand how your users retain every 3 days (bi-weekly). This grouping feature allows you to aggregate single-day data into larger intervals. It provides a way to view retention trends over a three-day span instead of looking at each day in isolation.
The documentation includes other specific questions that use day-based metrics. According to Mixpanel, one example asks what is the average video watch-time for a new subscriber in their first 30 days since subscription. This demonstrates how a 30-day window can be applied to measure engagement depth for new users. Another example according to Mixpanel asks how has my 7-day messaging retention changed over time. This shows how a 7-day metric can be tracked longitudinally to observe shifts in user behavior.
Before your health review, identify which day buckets your current chart uses. Note whether it tracks exact days like day 5 or grouped ranges like 15 to 30 days. This ensures you interpret the return periods correctly according to the specific rules defined by the tool.
Amplitude's starting event and return event
When you open the retention analysis builder in Amplitude, the tool requires you to define two specific data points to calculate the metric. The first is the starting event, which marks the initial user action. The second is the return event. Amplitude computes retention data by comparing the date of that starting event to the date of the return event you specified, according to Amplitude. This comparison determines whether a user counts as retained for that specific period.
The return event is the event that, when triggered, tells you a user is retained, according to Amplitude. You must select an event that represents meaningful re-engagement with your product. For a customer success manager, this might be a login, a feature usage event, or a support ticket interaction, depending on what defines active use in your context. The key is that the event must be one you can track and that reflects genuine user activity.
Amplitude allows you to select up to two return events for your retention analysis, according to Amplitude. This option lets you define retention more broadly. For example, you might count a user as retained if they either log in or use a specific feature. This flexibility is useful when a single event does not fully capture user engagement. However, you must be clear about which events you are using when you present the data in a health review. If you use two return events, state that explicitly so the audience understands the criteria.
When you present this data, note which return event or events you used. If you used two, list both. This transparency helps your stakeholders understand the basis of the metric. It also prevents confusion if they compare this data to other retention metrics that use different definitions.
Google's cohort example, kept as an example
Cohort reports analyze user retention over time by grouping users with a shared characteristic, according to Google. When you prepare for a health review, recognize that this grouping method defines the baseline for how return periods are measured. The documentation provides a specific scenario to illustrate this structure. For example, you could select the cohort of users that were acquired in the first week of September and follow that cohort for the next six weeks, according to Google. This example follows users acquired in the first week of September for the next six weeks.
The report response could show a weekly time series where say your app has retained 60% of this cohort after three weeks and 25% of this cohort after six weeks, according to Google. These figures serve strictly as an illustrative example within Google's own documentation. They do not represent a measured result for your specific customer base or your current health review. Do not treat the 60% or 25% figures as a benchmark you must match or a standard retention rate for your accounts. The purpose of these numbers is to demonstrate how a weekly time series might appear in the output.
When you read your own usage charts, distinguish between the structural definition of a cohort and the specific percentage values shown in examples. Google defines the cohort by a shared characteristic, such as acquisition time, according to Google. The example percentages describe a hypothetical app's performance over six weeks. Your health review requires you to look at your actual data, not the illustrative figures provided in this documentation. The weekly time series format is the key structural element to note, as it shows how retention is tracked at regular intervals.
Related guides
FAQ on retention labels before a health review
Which four period units does PostHog name?
PostHog names hours, days, weeks, and months as the four units for defining the retention period, according to PostHog. The retention period is defined by any number of hours, days, weeks, or months, according to PostHog.
What does PostHog say the first period always shows?
The first period always shows 100% because all users in that cohort must perform the event to be included.
How does Mixpanel describe day 5 retention?
Day 5 retention is the percentage of users who came back and did B exactly on the fifth day after the A event, according to Mixpanel.
What does Amplitude compare when it computes retention?
This comparison method focuses on the temporal distance between two specific user actions, according to Amplitude.
Sources
Limits of this post
This post is general operational information, not legal advice about notice periods or renewal clauses. It does not promise that any method keeps a customer. A review model checked the facts once, and that check can miss errors, so open the sources before you act.