How is revenue split across subscription, usage, and perpetual, and is usage growing faster than subscription?
№ 027 · Revenue mix by pricing model
Definition
MetricARR split across subscription, usage-based, and perpetual/other, with relative growth
Unit% of ARR per model; growth rate per model
Stack of subscription, usage, and perpetual revenue. Usage growing fast means capturing AI-product economics. Watch perpetual layer shrinking as monetization modernizes. Build by tagging each invoice line item with pricing model and aggregating per quarter. Example: subscription $7.8M, usage $4.8M (up from $0.8M), perpetual $0.6M.
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| Usage share shrinking as customers renegotiate to flat fees, or a usage line above half of ARR with no committed-spend floor, importing full revenue volatility | Roughly 6 in 10 SaaS companies now carry some usage-based component, most in hybrid form (base fee plus usage); pure-perpetual revenue is residual in modern SaaS books | Deliberate hybrid: a subscription base with a usage component that grows faster than the subscription line, mirroring the industry pattern (usage-based companies grew ~29.9% vs 21.7% for seat-based peers) |
B2B SaaS pricing-model adoption and growth data, 2021-2025 (OpenView usage-based pricing surveys, ~600-company benchmark; Metronome-linked datasets); AI products skew the recent mix toward usage · Adoption and growth-differential data are solid; actual revenue-mix distributions (what share of ARR is usage) are not published as quartiles, so the mix bands are structural readings. The trade-off the chart must show: usage grows faster and retains better but is less predictable quarter to quarter; hybrid with committed minimums is the industry's answer. KeyBanc 2025 notes companies monetizing AI still favor subscription models, so the drift is not one-way.
Category split omitted: Revenue-mix distributions by pricing model are not published as quartiles by two independent Tier 1-2 sources; only adoption rates and growth differentials exist.
When it looks bad
The usage band grows as a share of ARR while total ARR growth does not accelerate, meaning customers are shifting spend from committed subscription to variable usage rather than adding net-new spend: predictability falls with nothing bought in return.
Usage moves from 15% to 35% of ARR over a year; total growth stays at 18%. Forecast error on quarterly revenue doubles, the sales compensation plan breaks, and the board asks why the mix changed without the growth.
What to do about it
- Introduce committed-use floors as the usage share grows: annual commitments with drawdown pricing keep the expansion upside of usage (industry growth differential ~8-10 points) while restoring a forecastable base; multi-year commitments are anecdotally more common on usage contracts.
- Meter what tracks value, not what is easy to count: usage pricing on a unit customers cannot predict (raw API calls) creates bill anxiety and renegotiation to flat fees; unit selection is the difference between the 120%+ NRR pattern and churn.
- Ship usage transparency with the pricing: real-time dashboards, alerts at 50/75/90% of expected usage, and optional caps; surprise bills are the primary failure mode that pushes the mix backwards.
- For AI capabilities, price deliberately: KeyBanc data shows most AI monetization still lands as subscription; hybrid (platform fee plus metered AI actions) captures inference costs without giving up the committed base.
Sources
- The Founders Report (reporting OpenView usage-based pricing research) Usage-based companies grew 33.7% vs 23.2% (survey) and 29.9% vs 21.7% (600-company benchmark); UBP adoption ~45% of SaaS companies by 2021, rising thefoundersreport.com ↗
- Contrary Research Usage-based model companies showed NDR 125% vs broader SaaS index 114% (Jan 2023); AI cost structures pushing vendors toward usage models research.contrary.com ↗
- Metronome Hybrid models (base subscription plus usage) now common; subscription systems struggle with variable usage, driving billing re-architecture metronome.com ↗
- Culta (reporting OpenView data) 61% of companies with usage-based components run hybrid rather than pure consumption; UBP median NRR ~120% vs ~110% subscription-only culta.ai ↗
- KeyBanc Capital Markets and Sapphire Ventures 67% of companies already monetizing AI, tending to favor a subscription model for it sapphireventures.com ↗