Retention analysis chart: 3 steps across tools

The build order starts with data collection, not chart configuration. Amplitude states that you cannot use the Retention Analysis chart, or any other Amplitude chart, until you complete the instrumentation process, according to Amplitude. This prerequisite means the first action is verifying that your product events are being sent to the analytics platform. Without this step, the chart has no data to display.

Once instrumentation is complete, the next decision is selecting the events that define retention. Amplitude notes that the Retention Analysis chart helps you drive product adoption by showing you how often users return to your product after taking a specific action, according to Amplitude. You must identify one action as the starting event and another as the return event. Amplitude allows you to choose any event for both the starting and return events, according to Amplitude. This flexibility lets you define retention based on specific product behaviors rather than generic activity.

The remaining steps involve configuring the analysis parameters. After selecting the events, you add segments for each event to analyze them separately. You then set the interval and time range to define the periods you are measuring. Finally, you choose how the periods are labeled and read the resulting percentages. This sequence ensures that the chart reflects your specific definition of customer retention.

Write down the starting event and the return event your team would count as an active customer, then check that both are already being sent as events.

• Finish instrumentation first. You can't use the Retention Analysis chart, or any other Amplitude chart, until you complete the instrumentation process, according to Amplitude.
• Pick your events. When building a retention analysis chart, you can choose any event for both the starting and return events, according to Amplitude.
• Set the granularity. The granularity can be daily, weekly, or monthly, determining how the startOffset and endOffset are interpreted, according to Google.
• Check period labels. Period 0 is the default, with labels starting at 0, such as "Day 0" and "Week 0", according to PostHog.

Before you open the chart builder, confirm that your product events are actually being sent. According to Amplitude, you can't use the Retention Analysis chart, or any other Amplitude chart, until you complete the instrumentation process. This prerequisite means the first hour of your build session should be spent verifying that the specific actions you plan to track are firing in your data stream. If the instrumentation is incomplete, the chart will not function, so treat this as a hard gate rather than a parallel task.

Once you confirm the events are live, you need to select the two specific events that define retention for your customer. According to Amplitude, when building a retention analysis chart, you can choose any event for both the starting and return events. This flexibility allows you to define "retention" in terms that match your actual customer journey. For example, you might choose "Account Created" as the starting event and "Weekly Login" as the return event. The starting and return events can be the same action or different actions within your product.

If you are unsure which specific event best represents active usage, you have another option. According to Amplitude, you can also choose not to specify an event and instead tell Amplitude to use any active event. This setting is useful when you want a broad view of user engagement without narrowing it to a single feature. However, for a customer success manager trying to identify which specific behaviors correlate with long-term retention, choosing specific events is usually more actionable.

Once you have selected your starting and return events, the next phase involves defining how those events are grouped and measured over time. In Amplitude, each event has its own analysis and appears as a separate segment in the chart, according to Amplitude. If you are tracking multiple distinct user behaviors, you will need to create a segment for each one. To add another user segment, click + Add Segment and repeat steps 5 and 6, according to Amplitude. This allows you to compare different user groups or behaviors side-by-side within the same retention view.

The interval setting determines how the time periods are calculated. Different tools offer different granularities for this calculation. In Google Analytics, the granularity can be daily, weekly, or monthly, determining how the startOffset and endOffset are interpreted, according to Google. Choosing the right granularity depends on your product's usage cycle. If your customers use the product daily, a daily granularity might be too noisy, while a monthly granularity might hide short-term trends.

The time range defines how far back the chart looks. In PostHog, to analyze retention across a whole year, for example, you should configure your insight to say "in the last 12 months," according to PostHog. This setting ensures you have enough data points to see a stable pattern rather than just a snapshot of recent activity.

Finally, you can adjust how the time units are displayed. In Mixpanel, you can switch between time unit buckets, such as < 1 Day, Day 1 and Day 2, to see whether retention improves or declines along the curve, according to Mixpanel. Switching these buckets helps you identify if early drop-off is different from long-term churn.

Recommendation: Set your interval to match your product's natural usage rhythm. If your product is used daily, start with daily buckets. If it is used weekly, use weekly buckets. Record the specific interval and time range you choose in your chart settings sheet so you can reproduce the chart later.

The Retention - Docs - PostHog documentation outlines how to configure the initial period labels and the baseline for comparison. You can select between two retention reference options: "Period 0" (default) and "Previous period" according to PostHog. When you choose the default setting, labels start at 0, such as "Day 0" or "Week 0" according to PostHog. If your team prefers a one-based index, you can change the starting index to start at 1 instead of 0 according to PostHog. This adjustment lets you start the period labels at 1 instead of 0.

Understanding how the percentages are calculated is critical for interpreting the data correctly. By default, both reference options show the percentage of users who retained in each cohort relative to 'Week 0' according to PostHog. The percentages therefore use Week 0 as their default reference for each cohort. If you enable unbounded counting, the behavior changes significantly. When selected, the retention number for a specific time interval, say Day 2, includes users who returned on or after Day 2 according to PostHog. This implies that users who returned on Day 3, Day 4, and later are also counted in the Day 2 metric according to PostHog.

To apply this to your customer success workflow, decide whether you need to see immediate return behavior or cumulative return behavior. If you want to know how many customers came back specifically on Day 2, you should ensure unbounded counting is not selected, as that setting aggregates later returns into earlier days. To count customers who returned on or after Day 2, select unbounded counting. Both reference options show retention percentages relative to Week 0 by default. With "Period 0," the baseline is the initial cohort. The name "Previous period" does not mean its default percentage reference differs from Week 0.

The table below maps six specific chart settings to the vendor documentation that names them. Each row cites the publisher and the exact detail from their docs so you can verify the setting before you build.

Check each row against your current tool’s interface to confirm the option exists. If a setting is missing, note it in your chart settings sheet so you do not rebuild the chart later.

Record the specific choices you made during the build so the chart can be recreated without guessing. Amplitude states that you can't use the Retention Analysis chart, or any other Amplitude chart, until you complete the instrumentation process, so confirm that step is done before filling in the event names Amplitude. Create your own chart settings sheet to record the starting event, the return event, and the segment logic you applied. If you add another segment with + Add Segment, note the new event name and its interval here so the configuration remains clear for future reviews.

• Onboarding tools: compare 4 customer project plans
• Health score setup: 8 rules for weights and validity
• NPS survey configuration: 5 checks before launch

How do I build a retention analysis chart in Amplitude?

You must complete the instrumentation process before you can use the Retention Analysis chart or any other Amplitude chart, according to Amplitude. Once that prerequisite is met, you proceed to select your events and configure segments.

Can I use any event as the starting event in a retention chart?

Yes, you can choose any event for both the starting and return events when building the chart, according to Amplitude. This flexibility applies to both the initial action and the subsequent return action. You are not restricted to specific predefined event types for these selections.

Does PostHog start retention period labels at Day 0 or Day 1?

The default setting starts labels at 0, such as "Day 0" or "Week 0," according to PostHog. This is identified as the Period 0 option in the documentation. You can adjust the index if you prefer a different starting point, but the zero-based label is the standard default.

Which interval options does a Google cohort report allow?

The granularity can be set to daily, weekly, or monthly, according to Google. This selection determines how the startOffset and endOffset parameters are interpreted in the report. These are the three specific time unit buckets available for this configuration.

Does this chart setup apply to my own tool if I do not use Amplitude?

The specific steps described here are tied to the named sources and may vary by vendor. For example, the instrumentation prerequisite is specific to Amplitude, and the period label defaults are specific to PostHog. You should verify the requirements and defaults in your own tool’s documentation, as these settings are not universal across all analytics platforms.

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.