Which retention metrics matter in a commerce app?
Repeat purchase, cohorts, LTV, frequency, push, loyalty and churn signals — with definitions and an operating rhythm.
The most important commerce-app retention metric is not app opens but return to a business behaviour: repeat purchase, active loyalty use, saved product or a new cart. Measure retention by cohort and separately for new customers, returning customers and acquisition channels.
Separate product retention from revenue retention
A customer may return without buying, while some categories naturally have long purchase cycles. Report product activity, repeat purchase and retained revenue separately. Fashion can be monthly, books seasonal and supplements linked to a consumption cycle.
Metrics for a weekly review
Not every metric needs daily action. A weekly review should combine behaviour, purchase and communication channel.
- D7, D30 and monthly activity cohorts;
- repeat purchase rate and time to second order;
- orders per customer and purchase frequency;
- cohort LTV or margin when cost data is reliable;
- reactivation after push and email without confusing opens with purchase;
- loyalty usage and active-member share.
Churn signals and segments
A churn signal can be failure to return within the expected cycle, lower frequency, repeated zero-result searches, an abandoned cart or an unused reward. Thresholds must fit the category. A customer who buys every six months is not churned after thirty days.
From dashboard to experiment
Every observation should end with a hypothesis, segment, action and guardrail. Example: first-time buyers do not return within 45 days; test a reminder tied to the product cycle; measure second purchase and communication opt-outs. Without a comparison group, do not attribute the entire result to the campaign.
Retention interpretation traps
Retention can rise because of customer mix, seasonality or promotion rather than a new feature. Compare cohorts with similar source, country, first-purchase category and app version. An all-user average can hide both a strong loyal group and fast-churning promotional traffic.
Do not optimize one metric at the customer’s expense. Aggressive notifications can increase short-term opens while also increasing opt-outs or uninstalls. Every retention experiment needs a guardrail metric.
Owner operating rhythm
Analytics prepares cohorts and controls data quality, the product owner chooses the problem, CRM designs communication, and e-commerce owns the offer and margin. A weekly review ends with one or two experiments and an evaluation date. Monthly, review repeat purchase, LTV and consent opt-outs.
Sources
Firebase Analytics reports and audiences: https://firebase.google.com/docs/analytics/reports
GA4 user lifetime: https://support.google.com/analytics/answer/9947257
Amplitude retention analysis: https://amplitude.com/docs/analytics/charts/retention-analysis/retention-analysis-build
Frequently asked questions
- Is D30 retention right for every store?
- No. The window should reflect the category purchase cycle. D30 can work for app activity, while repeat purchase may require 60, 90 or 180 days.
- Does a push open count as retention?
- It is an engagement signal, not a business outcome. Follow the path to product, cart, purchase or loyalty use.
- How often should retention metrics be reviewed?
- Review operating signals and experiments weekly, and purchase cohorts and LTV monthly or in line with the category cycle. Daily alerts are mainly for data failures or sharp regressions.
Content updated: August 26, 2026
Google Preferred Sources
See GMI more often in Google
Add gmi.software as a preferred source. Google may highlight our new articles more often in Top Stories and supported AI experiences.