How to Build a Churn Prediction Dashboard in Amplitude
By the time a user cancels their subscription or stops logging in, the opportunity to save them has usually passed. Churn prediction shifts your focus from reacting to cancellations to proactively identifying users who are drifting away based on their behavior patterns. Amplitude's behavioral analytics make it possible to spot these signals early.
This guide shows you how to build a churn prediction dashboard in Amplitude that surfaces at-risk users, tracks leading indicators of disengagement, and helps your team intervene before users reach the point of no return. You'll combine engagement metrics, cohort analysis, and behavioral triggers into a single view that drives retention strategy.
Step-by-step guide
Define Your Churn Event and Timeframe
Before building anything, clearly define what 'churn' means for your product. For a SaaS tool, churn might be 'no login in 14 days' or 'subscription cancelled.' For a consumer app, it might be 'no core action in 7 days.' Create a custom event or behavioral cohort in Amplitude that captures this definition precisely — this becomes the target outcome your dashboard is designed to predict.
- Analyze your historical data to determine the inactivity window that most reliably predicts permanent churn
- Create a behavioral cohort for 'churned users' that you can use as a comparison group throughout the dashboard
Identify Leading Indicators of Churn
Use Amplitude's Event Segmentation and User Composition charts to compare the behavior of churned users versus retained users in the two weeks before churn occurred. Look for events that churned users performed significantly less (or more) than retained users. Common leading indicators include decreased login frequency, dropping core feature usage, increased support ticket submissions, and reduced session duration.
- Run a pathfinding analysis on churned users to see their last 10 events before churning
- Use the Compass chart to automatically identify events most correlated with retention
- Document the top 5-7 behavioral signals that differentiate churned from retained users
Build an Engagement Scoring Framework
Create a composite engagement score based on your leading indicators. Assign weights to each signal — for example, daily login (30%), core feature usage (25%), collaboration events (20%), content creation (15%), and settings customization (10%). Use Amplitude's formula feature in a custom metric to compute this score. Users whose score drops below a threshold enter the 'at-risk' zone.
- Calibrate thresholds by scoring your known-churned cohort and finding the score level that captured 80% of them before they left
- Create three risk tiers: Healthy (score above 70), At-Risk (40-70), and Critical (below 40)
Create the Churn Prediction Dashboard Layout
Build the dashboard with four key sections: a summary row showing total active users, at-risk count, and churn rate trend; a cohort retention chart showing weekly retention curves; a behavioral signals panel with charts for each leading indicator; and an at-risk user list with engagement scores. Use Amplitude's dashboard feature to combine these views into a single page that gives your team a real-time view of retention health.
- Add a retention analysis chart comparing this month's cohort to the previous three months
- Include a funnel chart showing the 'activation to habit' journey with drop-off points highlighted
Set Up Behavioral Cohorts for Risk Tiers
Create dynamic behavioral cohorts for each risk tier that update automatically. The 'Critical Risk' cohort should include users who have not performed the core action in the last 5 days AND whose session count this week is less than half their average. The 'At-Risk' cohort captures users trending downward but not yet critical. These cohorts feed into your intervention workflows and can be synced to tools like Braze or Intercom for automated outreach.
- Set up a cohort for 'Recently Churned' users to validate your prediction accuracy over time
- Create a 'Saved from Churn' cohort for users who were at-risk but re-engaged, to study what worked
Add Trend Lines and Comparative Analysis
Your dashboard should show not just current state but trajectory. Add week-over-week trend lines to your key metrics so you can see whether churn risk is increasing or decreasing across the user base. Include a comparison of engagement patterns between your at-risk cohort and your power users — the gap between these groups reveals which behaviors to encourage in your retention efforts.
- Add a stacked area chart showing the proportion of users in each risk tier over time
- Include a segment comparison breaking down churn risk by plan type, company size, or acquisition channel
Connect Alerts and Review Cadence
Set up Amplitude alerts that trigger when the at-risk cohort grows by more than 10% week-over-week or when the overall churn rate exceeds your target threshold. Schedule a weekly retention review where the product and customer success teams review the dashboard together, discuss which segments are trending poorly, and decide on interventions. The dashboard is only valuable if it drives action.
- Create a Slack alert for when high-value accounts (enterprise tier) enter the critical risk cohort
- Build an action log section in your team wiki that records what interventions were tried and their outcomes
Common mistakes
Using Lagging Indicators Instead of Leading Ones
Tracking metrics like 'subscription cancelled' or 'account deleted' tells you about churn that already happened. Your dashboard should focus on behavioral signals that predict churn 1-3 weeks before it occurs — decreased engagement frequency, reduced feature breadth, and declining session duration. By the time lagging indicators fire, the user is already gone.
Setting Churn Thresholds Without Historical Validation
Guessing that '7 days of inactivity means churn' without validating against your data leads to false positives that erode trust in the dashboard. Analyze your historical cohorts to find the actual inactivity window where re-engagement drops below 10% — that's your real churn boundary.
Building the Dashboard and Never Acting on It
A churn prediction dashboard that doesn't trigger interventions is just a fancy chart. Connect your at-risk cohorts to action workflows — automated emails, customer success outreach, in-app messages, or product changes. The dashboard should be a decision-making tool, not a reporting artifact.
Ignoring Segment Differences
Churn behavior varies dramatically by segment. A startup user might churn because they outgrow your tool; an enterprise user might churn because of a missing integration. Build segment-specific views in your dashboard so you can tailor interventions to the actual reasons each group disengages.
Tips
Start with your product's 'aha moment' metric as the core engagement signal — users who never reach it churn at the highest rates.
Use Amplitude's Lifecycle chart to visualize how users move between New, Current, Resurrected, and Dormant states over time.
Export your at-risk cohorts to your CRM or customer success tool weekly so the team can prioritize outreach to the highest-value accounts.
Track 'saves from churn' (users who were at-risk and re-engaged) to validate which interventions actually work and double down on them.
How Vantage helps
Vantage connects your analytics insights directly to product action. When your churn dashboard reveals a pattern, you can capture it as a signal in Vantage, generate a PRD to address the retention gap, and track the intervention from insight to shipped feature — closing the loop between data and product decisions.