Card-only subscription stacks structurally cap recovery around 55–65%. The remaining 35–45% is not a copy problem or a cadence problem — it's an instrument problem. Cards expire, get reissued, get blocked by risk models, and fail 3DS challenges. Bank rails and wallets don't fail the same way.
# What each instrument actually costs you
| Instrument | Failure rate | Cost per success | Recovery upside |
|---|---|---|---|
| Card (raw PAN) | 6–12% | 2.9% + $0.30 | Baseline |
| Card + network token | 4–8% | 2.9% + $0.30 | +15–30% |
| ACH (US) | 0.8–1.5% | $0.80 flat | +40% on high-ARPU |
| SEPA Direct Debit | 0.5–1.2% | €0.35 flat | +45% in EU |
| Apple Pay / Google Pay | 2–4% | 2.9% + $0.30 | +8–12% |
| PayPal | 3–6% | 3.5% + $0.30 | +5–10% |
| BNPL (Klarna, Affirm) | 4–7% | 3–6% | New-cohort acquisition |
The pattern: bank rails dramatically reduce involuntary churn at the cost of upfront friction. Wallets improve reliability marginally with almost no friction.
# When to prioritize which
- B2B annual, ACV > $5K: ACH/SEPA. Card fees alone justify migration.
- B2C monthly, ARPU $10–50: Apple/Google Pay as primary, card as fallback.
- LATAM/India consumer: Pix, UPI, local wallets. Cards are the fallback, not the default.
- Enterprise annual: Wire + invoice. Cards should not be the primary instrument above $25K.
# The fallback instrument pattern
The highest-recovery pattern we've measured: collect a secondary payment method at signup. When the primary fails and dunning step 2 hasn't resolved in 48 hours, silently attempt the secondary. Recovery lifts 12–18% incremental because the customer doesn't need to act at all.
Legal caveat: mandates matter. SEPA and ACH require explicit authorization for each instrument. Read your PSP's fallback documentation before shipping.
# Migration paths that don't break billing
Introduce new instruments as new plans, not retroactive changes:
- Launch "Pro Annual (ACH)" as a distinct SKU.
- Offer 2–5% discount as the incentive to switch.
- At renewal, prompt existing card customers to migrate.
- Keep card as the universal fallback.
Don't force-migrate. Don't remove card. Both destroy trust and increase voluntary churn far more than involuntary churn declines.