Retention, conversion, and cohorts

A cohort groups customers or subscription journeys by an anchor date and follows later outcomes. Always identify the anchor, denominator, observation window, and maturity before comparing two cohort charts.

Subscription retention

The anchor is the subscription's first funded service period, not the beginning of a free trial. Cohorts are grouped by paid-start month and filtered to weekly, monthly, or yearly product duration. Catalog package type supplies duration when available; bounded observed service length is the fallback for older data.

Each renewal is measured against the original paid cohort. A product/duration change starts a new journey. A same-product resubscribe after a gap stays in its original journey. Grace time alone is not a funded renewal, and refunded/revoked periods do not count as retained.

Relative cells remain empty until every subscription in the cohort has reached that renewal opportunity. Do not read an immature blank as 0% retention. Absolute mode shows the corresponding counts; relative mode divides by the original cohort.

Trial conversion and cancellation

Trial conversion cohorts start at the trial's own beginning. Overlapping observations of the same trial collapse into one trial. A funded, non-refunded period must start at the trial boundary to count as direct conversion; a later resubscribe is not attributed back to that trial.

Trial starts reconcile into converted, open, and abandoned. Open trials make a cohort incomplete. Trial cancellation is narrower than abandonment: the trial must end without conversion after an auto-renewal opt-out inside the chosen window. An opt-out that was reversed before the user paid is not a cancellation.

Cancellation windows are 1, 2, 5, or 7 days, or unbounded. They are measured as N × 24 hours from the trial timestamp, not N UTC date labels. If the user opted out repeatedly, the last in-trial decision matters.

Customer conversion

Initial conversion and conversion to paying use the customer's first-seen date as the anchor. First seen means the CherrySub customer record, not every app install or account registration in your own system. Minting a customer token alone does not add a customer to this denominator. Windows are Day 0, 3, 7, 14, 30, or unbounded.

Initial conversion includes trial starts and product purchases. Conversion to paying requires a funded transaction; a free trial alone is not paying. Store/environment/product filters affect the conversion numerator, not the new-customer denominator. A recent cohort can still be maturing.

Refund timing matters: a refund removes paid conversion only from windows in which the refund occurred. Do not compare a seven-day conversion window to an unlimited lifetime outcome as if their samples were identical.

Realized lifetime value

Both LTV reports use first-seen customer cohorts and preserve each currency separately. A lifetime window is Day-0 inclusive: Day 7 includes eight UTC dates, Day 0 through Day 7. Available windows are 0, 3, 7, 14, 30, 60, 90, 180, 365 days, or unbounded.

Revenue includes purchase value in that lifetime and subtracts refunds whose timestamps fall within the same lifetime. Per-customer LTV divides by every new customer. Per-paying-customer LTV divides, per currency, by customers with net-positive funded value in that lifetime. Filters do not silently shrink the all-customer denominator.

This is realized observed value, not a forecast of future lifetime revenue. Taxes, commission, and currency conversion are not inferred.

Cohort matrix

Choose one anchor: new customers, initial conversions, or new paying customers. A conversion anchor is the first qualifying event over the customer's history, then filtered into the requested cohort range. A first purchase that was refunded does not by itself establish an unrefunded paying anchor.

Month 0 is the cohort's own calendar month. This differs from subscription retention's renewal-opportunity columns. Measures are revenue, cumulative realized LTV, realized LTV per customer, or retained subscriptions. Choose an actually earned currency for monetary measures.

Revenue/refunds are assigned to the corresponding purchase period in this matrix; cumulative LTV builds from those periods. Cohort size remains the denominator over time. Retained subscriptions count unique chains at period end. Future/unlived periods remain absent from totals.

Interpret the result

Before reporting a change, record the anchor, selected product/store/environment, currency, lifetime/window, and maturity. A shifted denominator, incomplete cohort, or different refund attribution can explain a difference without a behavioral change.

For a customer-level example, trace the first-seen, trial, purchase, and refund dates through Transactions and events. For every available metric name and formula, use Metric definitions.