Every subscription operator wants to know one thing: is my involuntary churn normal? Here are the answers, cut across industry and plan value, with actual percentiles instead of "somewhere between 5 and 15%."
Numbers below are aggregated from RRLabs' anonymized production data across ~800 workspaces in H1 2026, cross-checked against public disclosures and payment provider reports.
# Involuntary churn rate (% of active MRR lost per month)
| Industry | p10 (best) | p50 (median) | p90 (worst) |
|---|---|---|---|
| B2B SaaS (annual-heavy) | 0.3% | 0.7% | 1.6% |
| B2B SaaS (monthly-heavy) | 0.6% | 1.4% | 3.2% |
| Consumer subscriptions | 1.1% | 2.5% | 5.8% |
| Streaming/media | 0.9% | 2.0% | 4.5% |
| E-commerce subscription boxes | 1.3% | 3.1% | 6.9% |
| Fintech consumer subscriptions | 0.8% | 1.9% | 4.2% |
| Gaming subscriptions | 1.2% | 2.7% | 5.9% |
The pattern is consistent: plans billed monthly have 2–3x higher involuntary churn than plans billed annually, and consumer categories have 2–4x higher involuntary churn than B2B.
# Failed-charge rate (% of attempted recurring charges that fail)
| Category | Median failure rate |
|---|---|
| B2B SaaS on cards | 4.2% |
| B2B SaaS on ACH/SEPA | 1.1% |
| Consumer, developed markets | 7.8% |
| Consumer, emerging markets | 12.5% |
| Digital wallets (Apple Pay, Google Pay) | 2.9% |
Cards remain the dominant instrument, and the dominant source of failure. Every 100 basis points of failure-rate improvement — through smart retries, network tokens, and updater programs — is worth roughly 60bps of MRR retained.
# Day-7 recovery rate (% of failed dollars recovered within a week)
| Segment | p10 | p50 | p90 |
|---|---|---|---|
| No dunning system | 8% | 14% | 22% |
| Basic Stripe Smart Retries | 18% | 27% | 38% |
| Custom dunning cadence | 26% | 38% | 51% |
| AI-driven recovery (RRLabs et al.) | 34% | 47% | 62% |
The gap between "we use Stripe defaults" and "we invested in recovery" is roughly 20 percentage points of recovered dollars. On a $10M ARR business with a 7% failure rate, that's ~$140K/yr recovered.
# Time-to-recovery
| Metric | Best-in-class | Typical |
|---|---|---|
| p50 time-to-recovery | Under 12 hours | 2–3 days |
| p90 time-to-recovery | Under 3 days | 7–10 days |
Fast money is worth more than slow money for two reasons: it reduces the exposure window (customer churn probability rises with time), and it improves working capital.
# What drives the spread
The delta between p10 and p90 within each category is enormous — often 4–5x. What consistently separates the best from the worst:
- Segmented cadences by failure code and plan value (not one cadence for everyone).
- Multi-channel (email + WhatsApp/SMS in the right markets).
- Dedicated recovery sending domain with clean DMARC.
- Real reply monitoring — the top decile responds to inbound replies inside 4 business hours.
- Update-payment friction under 3 taps from any message.
# What doesn't matter as much as people think
- Cadence length beyond 4 touches.
- Discounts / offers in dunning mail (small lift, meaningful margin hit).
- Countdown timers (short-term uplift, long-term trust damage).
- Aggressive retry frequency (issuer relationship damage).
# What to do with these numbers
Pick your row. Look at your own equivalent metric. If you're at p50 in your category, moving to p10 is worth writing down as a number and staffing accordingly. If you're at p90, moving to p50 is usually a matter of weeks, not quarters.
Methodology note: RRLabs anonymized production data, H1 2026, excluding workspaces with less than 90 days of history or less than $10K in monthly recurring charges. Public data cross-checks from Stripe, Adyen, and Braintree annual reports.