How do we decide whether to migrate ALL subscribers at once versus staggering the cutover by cohort (e.g., newest subscribers first, as a lower-risk test group)?

When does staggering by cohort beat migrating every subscriber at once?

Staggering by cohort beats migrating every subscriber at once above roughly 2,000 active subscribers, or whenever cards must be re-collected. A first cohort turns an unknown failure rate into a measured one, so the rest of the base moves on evidence. Once the cost-benefit case for leaving Recharge or Skio says go, four criteria set the cutover's shape, in the order they decide it.

1. Subscriber count. Below roughly 2,000 subscribers on simple plans, Deploi's migration record says buy the destination's migration service (October 2026), and a single window is fine. Above that line, stagger.
2. Vault portability. If cards must be re-collected, stagger at any size, because you are running a comms campaign, not a data move, and campaigns need a test cell.
3. Billing-date spread. Anniversary billing gives you natural cohorts. Fixed-date billing does not, so a stagger means splitting a single billing event, which is harder.
4. Rollback appetite. A staggered cutover keeps rollback cheap for the first cohort. A big-bang migration's rollback is a second migration.

Does your destination's migration process allow a cohort at all?

Your destination's migration process decides whether a cohort is possible, so check it before you plan one. Shopify recommends pausing billing in the current app while payment methods and contracts move, to avoid double-billing (Shopify developer docs, October 2026). In a staggered cutover that pause applies one cohort at a time, so the old app must stop billing some subscribers while it keeps billing the rest. The full order of a subscription migration has its own answer.

Which subscribers belong in the first cohort?

The first cohort should be your newest subscribers on cards your current processor holds, leaving out anyone on a grandfathered price or a bespoke discount. That group has the freshest credentials, the least tenure at stake and the fewest odd contract fields, so a failed transfer costs least there.

Cards the destination can reach without re-collection make the cleanest test. Shopify can migrate saved payment methods from Stripe, Braintree, PayPal Express and Authorize.net, which keep billing existing contracts as a secondary gateway. That token migration "is available on all Shopify plans" (Shopify Help Center, October 2026). Moving the card numbers themselves needs Shopify Plus or Enterprise. Leave grandfathered subscription prices for a later cohort, because each contract line carries its own price and a mapping error there is a price change. Size the cohort to produce a readable transfer rate: a few hundred contracts, not a dozen.

What should you read before moving cohort two?

Three read-outs gate cohort two (Deploi practice). If any one is outside expectation, stop and fix rather than proceed more carefully, because cohort two inherits every defect cohort one exposed.

• Payment-method transfer rate, measured seven days after the cohort's cutover. Measure it the way the card-retention benchmark you write into a migration agency's scope does.
• Charge accuracy on the first billing event: right date, right amount, charged once.
• Support tickets per thousand subscribers migrated, against your pre-migration baseline.

Take all three read-outs after cohort one's first billing event has run, because charge accuracy and ticket volume only show once a charge has gone out. Cohort two waits for them, and the calendar that wait adds belongs in the migration's own risk budget.

When should every subscriber move in one window?

Every subscriber should move in one window when a stagger is impossible, or when it costs more than the risk it retires. A stagger costs two platforms, two sets of subscriber emails and two support scripts for as long as it lasts, with each subscriber billed and emailed by one platform only.