The five axes
Tenure
How long the subscriber has been continuously paying. 12+ months on a renewal cycle behaves very differently from a 3-month new subscriber. Long-tenure cohorts often migrate at higher rates.
Geo
Primary markets vs secondary. Compliance and reachability vary by region; payment-method support varies (e.g., SEPA, Brazilian Pix). Sequencing by geo lets you start where the path is clearest.
Engagement
Quintiles of in-app activity. High-engagement subscribers are typically easier migrations (they care about the app); low-engagement subscribers carry more churn risk on any touchpoint.
Renewal window
Time until next renewal. Subscribers near a renewal cycle tend to be more receptive — the renewal is on their mind. Longer-window subscribers can wait, and waiting compounds the recovery.
Plan
Annual vs monthly, premium vs basic, trial vs paid. Annual cohorts have different conversion mechanics than monthly (card-save vs early-charge); premium tier subscribers often have higher migration intent.
How the axes get used
The pilot samples across all five axes at once — a stratified slice of the base with every cohort represented, each measured separately so response can be attributed per cohort rather than averaged away. Two to three waves build a per-cohort read before the program scales. Migration waves then run stratified-parallel — every cohort starts in week one at a small allocation, and volume ramps wherever that cohort’s benchmarks clear. Each wave’s cohort spec tightens or widens based on what the prior waves showed.Ongoing qualification
The five axes scope the project-phase migration. After that, the migration extends itself: new mobile cohorts continue to qualify as they cross the same thresholds the pilot validated. When a new cohort crosses the qualified band, the standardized migration sequence runs — branded email, web checkout that preserves the subscriber’s existing economic terms, entitlement bridge — without per-cohort tuning. The thresholds were set during the pilot; new cohorts cross them as time and engagement accumulate.Why five and not more
There are more axes you could split on (signup source, device tier, language, billing currency, etc.). Five is the count where:- The combinatorics stay tractable. Beyond five, the cohort space explodes and starves any individual cell of statistical power.
- The axes are independently meaningful. Tenure, geo, engagement, renewal window, and plan all carry distinct migration-rate signal. Adding a sixth typically correlates with one already on the list.
- The reporting fits on one screen. A wave summary that’s one screen tall keeps the gating decision fast; a wave summary that scrolls is a slower decision.