How does net revenue retention differ across pricing models, and what does each model imply for predictability?
№ 028 · NRR by pricing model
Definition
MetricNet revenue retention split by pricing model (usage-based vs subscription)
Unit% of year-ago revenue by model
Three NRR lines (subscription, usage, perpetual). Usage NRR is high but volatile, subscription steady, perpetual at or below 100%. Build from cohort billing segmented by pricing model. Mix shift toward usage shows up as blended NRR rising. Example: subscription 108%, usage swings 118-145%, perpetual 92%.
- NRR
- Net revenue retention. Cohort revenue this period divided by the same cohort revenue a year ago, including expansion, contraction and churn. New logos excluded.
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| Usage-based below 105% (usage pricing without usage growth is churn with extra steps); subscription below 100% | Usage-based ~120-125% vs subscription ~110-115% in survey data; consumption-model top quartile 122% vs 110% (usage tiers) and 109% (no usage) in earlier benchmark vintages | Usage-based cohorts 125%+; subscription cohorts 115%+ |
B2B SaaS by pricing model, 2021-2023 survey vintages (OpenView UBP surveys, Metronome-linked data); the differential has held across vintages even as absolute NRR compressed market-wide · The 10-15 point usage premium comes from expansion occurring without a sales cycle, but it is symmetric: usage NRR falls faster in downturns as consumption contracts (the 2022-2023 cloud optimization wave hit consumption models hardest). Definitions vary on whether committed-but-unused spend counts; state your treatment. Data vintages predate the AI-native wave, whose low-price tiers show much weaker retention in ChartMogul data.
| Category | Bottom 30% | Median | Top 30% |
|---|---|---|---|
| By pricing model | usage-based under 105%; subscription under 100% | usage-based ~120-125%; subscription ~110-115% | usage-based 125%+; subscription 115%+ |
When it looks bad
The usage-model bar swings across periods while the subscription bar holds: usage NRR prints 130% one year and 98% the next as customer consumption cycles, revealing that the premium was market beta, not product-driven expansion.
Usage cohort NRR: 128% (expansion year), 96% (optimization year); subscription cohort: 108% and 105% across the same two years. Blended NRR whipsaws 6 points and forecasting credibility with the board goes with it.
What to do about it
- Split usage NRR into volume effect and unit-price effect: retention driven purely by customers consuming more of the same unit collapses when they optimize; add cross-sell modules and committed tiers so part of the expansion is structural.
- Convert volatile usage accounts to annual commitments with drawdown at renewal, trading a few points of peak NRR for a floor; the 2022-2023 consumption-optimization wave is the case study for why the floor is worth it.
- Do not assume the usage premium transfers to low-price AI products: ChartMogul late-2025 data shows sub-$50/month AI plans retaining catastrophically (near 32% NRR); at low ACV, packaging and activation dominate pricing model.
- Report NRR by model with a consumption-adjusted view (normalizing for customers' own volume cycles) so the board can distinguish your expansion execution from your customers' business cycles.
Sources
- The Founders Report (reporting OpenView State of UBP, 2nd edition) Usage-based companies at 125% NRR vs 115% for subscription peers; public UBP companies ~25% higher NDR and 38% faster growth thefoundersreport.com ↗
- Contrary Research 2021 vintages: top-quartile NDR 122% for consumption-primary vs 110% usage-tier and 109% no-usage; UBP NDR 125% vs 114% SaaS index (Jan 2023) research.contrary.com ↗
- Culta (reporting OpenView benchmark data) UBP median NRR ~120% vs ~110% subscription-only; expansion accrues automatically with product adoption culta.ai ↗
- Benchmarkit NRR benchmarks filterable by pricing model, enabling within-model comparison benchmarkit.ai ↗
- Digital Applied (reporting ChartMogul data) AI-native retention weak at low price points: sub-$50/month plans near 32% NRR vs 85% above $250/month in late-2025 ChartMogul data, a caution against assuming the usage premium transfers to AI-native low-ACV products digitalapplied.com ↗
- KeyBanc Capital Markets and Sapphire Ventures Companies monetizing AI tend to favor subscription models, trading expansion upside for predictability sapphireventures.com ↗