Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Fintech cohort retention is the share of a defined group of users who return to perform a meaningful action after joining that group. To calculate it accurately, specify the cohort-entry event, return event, identity and deduplication rules, time interval, timezone, eligibility, and treatment of churn or reactivation. “Day 7” can mean activity on that interval alone or activity on that interval and later; those are different metrics.
Define the cohort and the behavior that counts as retention
A cohort is a group of people assigned according to a shared entry event and time—for example, customers who first funded an account in the same week. The entry event should reflect the point at which the analysis begins, not merely whichever event is easiest to track.
As an Amazon Associate I earn from qualifying purchases.
Choose a return event that signals ongoing value from the fintech service. A login may be useful for diagnosing product use, but it does not necessarily show that a customer received value. A payment, transfer, investment, loan repayment, or policy-related action may be more relevant, depending on the product and its promised usage cycle.
Recommended Free Tools
Before publishing a result, document these choices:
#1 Best Overall
- Entry event: What action places a person in the cohort, and when does cohort membership begin?
- Return event: Which later action counts as retained, and why does it represent meaningful engagement?
- Unit and identity: Is the unit a person, account, device, or subscriber? Which stable identity joins the entry and return events?
- Interval and boundaries: Does “Day 1” mean the first 24 hours after entry or the next calendar date?
- Eligibility: Must a person have had the full interval in which to return before being included?
- Re-entry and reactivation: Can a person enter more than one cohort, and how are returning former customers treated?
These are not reporting details to leave implicit: changing any of them can change the number.
Calculate exact-interval and on-or-after retention separately
Return On: activity in the specified interval
For a specified interval X, exact-interval retention is:
Return On retention at X = unique users from the cohort who performed the return event in interval X ÷ unique users who entered the cohort
For example, Day 7 Return On counts cohort members who return during Day 7 only. A person who returns on Day 6 but not Day 7 does not count in the Day 7 numerator.
Rank #2
Return On or After: activity in that interval or any later one
On-or-after retention uses the same entry denominator, but counts a person if they perform the return event in interval X or later:
Return On or After retention at X = unique users from the cohort who return in interval X or later ÷ unique users who entered the cohort
A user who returns on Day 10 can therefore count in Day 7 Return On or After, even though they did not return on Day 7. Amplitude documents Return On and Return On or After as distinct retention measures; its overall on-or-after view includes only start-event cohorts old enough to have reached the interval. A cohort row uses that cohort’s fixed entry count as its denominator. See Amplitude’s retention calculation documentation.
Do not compare these measures as if they answered the same question. Return On measures whether activity happened at a particular point; Return On or After measures whether users came back by or after that point. Label the method beside every chart, table, and comparison.
Keep identities and denominators consistent
Use unique people (or another explicitly chosen unit) in both numerator and denominator. Counting return events in the numerator while counting unique entrants in the denominator can produce a misleading rate, especially when users can act repeatedly. If your product supports multiple accounts per person, decide whether retention is person-level or account-level and apply that rule throughout.
For an interval-level result, exclude cohorts that have not had enough observation time, or clearly mark the interval as incomplete. Otherwise, apparent differences may reflect exposure time rather than customer behavior.
Choose a time model and apply it consistently
Rolling elapsed-time intervals
With rolling intervals, time is measured from each person’s cohort-entry timestamp. Day 0 begins at the start event; Day 1 covers hours 24 through 48 after it. Two people entering on different dates therefore have different Day 1 windows.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Calendar intervals
Calendar-day retention assigns events to dates in the selected project timezone. The same event timestamp can fall on different calendar days under different timezone settings. Weekly results can also change depending on which day begins the week.
Amplitude documents both rolling 24-hour windows and strict calendar-date reporting, with calendar intervals tied to project time settings. See Amplitude’s explanation of time in retention analysis and its guide to interpreting retention results. Select the convention that fits the question, then preserve it across dashboards and comparisons. Check boundary timestamps directly, including timezone conversion and daylight-saving transitions where relevant.
Do not confuse behavioral retention with subscription churn retention
Behavioral retention asks whether someone performed a chosen return event after entering a cohort. Subscription retention can instead ask whether a paid subscription remains active, or whether it has churned, at a point in time. These definitions describe different outcomes and should not share an unlabeled “retention” figure.
For example, Stripe’s Billing cohort method assigns subscribers according to when they first started generating positive monthly recurring revenue (MRR) from active paid subscriptions, then measures the percentage that has not churned by month end in UTC. Resubscribers remain in their original cohort. That is not the same as measuring whether a person used a banking or investing feature in a later interval. See Stripe’s description of subscriber cohorts and cohort retention in Billing.
For either approach, state how reactivation is handled. A behavioral measure might count a user who returns after a period of inactivity; a churn-based subscription measure may treat a former subscriber who resubscribes according to its cohort rule. The rule must be explicit to make the result interpretable.
Best Value
Validate the number against events and identities
- Inspect event definitions. Confirm that the start and return events represent the intended behavior. Check for duplicate client/server emissions, missing events, and late-arriving records.
- Verify identity resolution. Confirm the key used to deduplicate people and join events. Reconcile unique cohort entrants and returners with event-level records.
- Check cohort membership. Test whether repeated start events accidentally add one person multiple times or create multiple cohort memberships. Document any intentional re-entry rule.
- Test time boundaries. Manually inspect sample timestamps around interval edges, timezone changes, week starts, and daylight-saving transitions.
- Check maturity. Flag or exclude intervals that cohorts have not had time to complete, and compare cohorts at the same age.
- Hand-calculate a small cohort. List its user IDs and event timestamps, apply the stated interval rule, and reconcile the result with the analytics output.
- Reconcile chart aggregation. Check whether an overall result pools users across cohorts or averages cohort percentages, and document the method.
- Segment carefully. Compare relevant groups such as acquisition channel, product type, or customer state only when the definitions and cohort sizes remain meaningful.
Retention curves can be counterintuitive when they include incomplete periods. Amplitude notes that an ongoing curve may rise at later intervals because only cohorts old enough to reach those intervals are included. It also notes that deduplication and incomplete periods can make cohort-row sums differ from totals. Validate the behavior of the reporting system rather than assuming its chart semantics match another tool’s.
Interpret aggregate rates and comparisons cautiously
A pooled rate of total retained users divided by total cohort entrants is not generally the same as the arithmetic mean of cohort-level percentages. They match only when the cohort sizes are equal. A pooled rate gives larger cohorts more influence; an unweighted mean gives every cohort the same influence. State which one the chart uses, and check how the analytics platform aggregates its result.
Before comparing two retention numbers, align the definitions and context:
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches- Return On, Return On or After, or churn-free subscription retention.
- Entry and return events, identity key, deduplication, and re-entry policy.
- Cohort age, maturity, observation window, and interval length.
- Timezone, calendar-day or rolling-time convention, and week-start setting.
- Aggregation method, cohort size, and segment composition.
- Product lifecycle and business model: a daily-use trading product may have a different natural cadence from monthly billing or insurance.
Amplitude’s fintech guide recommends examining onboarding behavior, drop-off, and feature-engaged groups as ways to investigate retention patterns. Such comparisons can suggest where to look, but association between feature use and retention does not by itself show that the feature caused retention. See Amplitude’s fintech guide.
Do not treat illustrative chart values as fintech targets
The sources cited here do not establish a universal fintech retention benchmark. Example percentages shown in analytics documentation illustrate how a chart works; they are not industry targets. Set internal comparisons around a clearly defined metric, comparable cohort ages, and the product’s actual customer lifecycle rather than borrowing an unlabeled target.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
