Define the question before entering numbers
Choose a start date, end date and membership definition. For example, you may want to know what proportion of people with an active paid membership on 1 January still have that status on 31 March. Save the list of unique members at the start. That fixed list is your cohort; it does not expand when new people join.
The denominator is the number of people in that list. The numerator is the number of those same people who meet the documented end condition. Count people once, even if they have multiple products. If you instead want to measure contracts or revenue, name that measure separately so it is not mistaken for member retention.
Worked fictional example: 200 starting members
Retention is 180 ÷ 200 × 100 = 90%. The 30 new members increase the ending total, but they were not at risk of leaving throughout the same starting period. Using 210 ÷ 200 would produce 105%, which describes neither the survival of the starting cohort nor a valid retained proportion.
In this simple example, 20 ÷ 200 × 100 = 10% of the cohort does not meet the end definition. That is the complement of retention under the same rules. Do not label it a separate churn measure unless its membership definition and period also match.
| Count | Members | Treatment |
|---|---|---|
| Starting cohort | 200 | Denominator |
| Starting members meeting the end definition | 180 | Numerator |
| Starting members not meeting the end definition | 20 | Outside the retained group |
| New members joining after the start | 30 | Excluded from this cohort calculation |
| Possible ending total | 210 | 180 retained plus 30 new; not the numerator |
Decide how pauses and returners count
There is no single pause policy that fits every membership model. You could define a continuing contractual membership as retained even during an approved payment pause, or report active paying membership with pauses shown separately. Either approach needs an explicit label. Changing the rule between periods can create an apparent improvement without any change in members' behaviour.
For an end-date measure, a starting member who leaves and rejoins before the end may qualify as retained under your documented policy. That does not mean they stayed continuously. If uninterrupted membership is the question, use a separate continuous-retention definition and track the intervening dates. A returner who was not in the starting cohort is never added to this cohort's numerator.
- State the status required on the start date and the end date.
- Record whether approved pauses remain eligible and whether their fees matter.
- State how cancellation notice periods, expiry dates and reactivations are treated.
- Keep a separate count of pauses and returns when these materially affect interpretation.
Calculate and check the result
- Freeze the starting list using stable member identifiers and remove duplicates.
- Match that list against the end-date records. Do not rely on the total member count alone.
- Apply the same documented eligibility rules to each starting member.
- Count the eligible end members and divide by the starting count, then multiply by 100.
- Check that both counts are whole numbers, the starting count is greater than zero and retained members range from zero to the starting count.
- Save the dates, counts, definition and any data corrections alongside the percentage. Round only the displayed result, retaining enough precision for later comparison.
What retention can and cannot explain
The result describes the chosen cohort over the chosen interval. Monthly and annual retention cover different opportunities to leave; do not compare them as equivalent percentages. Compare like periods and membership definitions, and examine new versus established members separately when that answers an operational question.
A small cohort can move sharply when one person leaves: one member out of ten changes the result by ten percentage points. Check the underlying count before treating a percentage change as a trend. Retention does not reveal attendance, satisfaction, revenue or why people left. Investigate those separately, and do not attribute a change to coaching from this calculation alone.
Common calculation errors
- Using ending total minus new joiners without checking identities. Transfers, duplicate records and reactivations can make that shortcut unreliable.
- Counting visits instead of members. One frequent visitor still contributes one person to the cohort.
- Moving the start date after seeing the result. Choose the period first and keep comparable reports.
- Reporting a percentage for an empty starting cohort. With a zero denominator the rate is undefined; report no eligible starting members.
- Treating an industry percentage as a target without checking its period and definitions. This guide does not propose a universal retention benchmark.
Practical checklist
- Write the start and end dates.
- Define eligible membership and the unit: unique people.
- Save the starting cohort.
- Document pauses, expiry, notice periods and returners.
- Match the same people at the end.
- Exclude all later joiners.
- Validate counts and show numerator and denominator.
- Compare only reports with compatible definitions.
Questions about this topic
Can retention exceed 100%?
Not for this starting-cohort member measure. The numerator is a subset of the denominator. A result above 100% usually means that new joiners, duplicate people or a different unit have entered the numerator.
Can I use the cancellation count instead of matching members?
Only if it fully reconciles with the same starting cohort and definition. A total cancellation list can include people who joined during the period, while expiry or status changes may be recorded elsewhere. Matching the cohort makes the question explicit.
Should a member on a freeze count as retained?
Choose a policy that matches the question, document it and label the report. Continuing contractual membership and active paying membership answer different questions. Show freezes separately if they could obscure a change.
What period should I choose?
Use a period suited to the decision, such as a monthly service review or a longer membership review. Keep dates and definitions consistent. If you change the interval, establish a comparable baseline rather than comparing the new rate directly with the old one.