Subscription Metrics
Date: 2026-08-16
Recurring revenue changes the arithmetic — churn becomes an explicit event and lifetime value becomes computable. It also introduces a metric nobody watches: a large share of cancellations are failed payments rather than decisions.
What it is
Subscription metrics describe a business with recurring billing, where revenue continues until an explicit cancellation or payment failure.
The core set:
| Metric | Is |
|---|---|
| MRR | Monthly recurring revenue — active subscriptions × monthly value |
| Churn rate | Proportion of subscribers lost per period |
| Net revenue retention | Revenue from a cohort this month ÷ last month, including upgrades and downgrades |
| ARPU | Average revenue per user |
| Subscriber LTV | Monthly contribution ÷ monthly churn rate |
The MRR movement view
More useful than the MRR total, because it shows what’s driving the number:
opening MRR £120,000
+ new £14,000
+ expansion (upgrades) £3,500
− contraction (downgrades) −£2,100
− churn −£9,800
─────────
closing MRR £125,600
Growth of £5,600 on £14,000 of new business. Three-fifths of new revenue is replacing losses — a treadmill invisible in the headline total. This decomposition is the single most useful subscription report.
LTV is violently sensitive to churn
At £15 monthly contribution:
churn 3% → £500
churn 5% → £300
churn 8% → £188
A two-point change in churn moves LTV by 40%. Which means any subscription LTV figure quoted without its churn assumption is close to meaningless, and small churn improvements are worth more than equivalent acquisition gains — Customer Lifetime Value.
The metric nobody watches
Involuntary churn — subscriptions ending because a payment failed, not because anyone decided to leave. Expired cards, insufficient funds, bank declines.
It’s frequently a substantial share of total churn and it’s the cheapest to fix, because these customers weren’t leaving. Treating it as ordinary churn means running retention campaigns at people whose only problem is a card that expired — Involuntary Churn and Dunning.
Always split churn into voluntary and involuntary before analysing it. They have different causes and different fixes.
Retail subscriptions differ from SaaS
Worth flagging, because most subscription metrics literature is SaaS:
- Physical goods carry COGS every cycle. Revenue retention doesn’t mean margin retention if product costs rise
- Skipping is not churning. Replenishment subscriptions allow pauses and skips; counting a skip as churn overstates it badly
- Fulfilment cost recurs, so contribution per cycle is well below revenue per cycle
- Expansion revenue is limited. There’s no upgrade tier in a consumables subscription, so net revenue retention above 100% is rare — a SaaS benchmark that doesn’t transfer
Measuring it
- Cohort churn, not blended. Blended churn falls automatically as the base ages, because long-tenured subscribers churn less — Cohort Analysis
- Churn by tenure. The first two or three cycles are where most losses happen; a blended rate hides that the fix belongs in onboarding
- Contribution, not revenue — Contribution Margin
- Pauses and skips separately from cancellations — Cancellation Flows
- Failed-payment recovery rate as a standing metric
Related: Retention and Churn for the non-subscription case, where churn has to be defined rather than observed.