"How does usage-based pricing affect net revenue retention?"

"It can raise NRR when customers expand usage. Snowflake reported 158% in its 2020 S-1 and 125% in Q4 fiscal 2026, and Twilio reported 140% for 2018. The same mechanism lowers NRR when customers optimise spend, so NRR tracks customer activity as well as product quality."

"How does usage-based pricing change CAC payback?"

"The first invoice is typically smaller, so payback depends on expansion rather than contract value. Model payback on a ramp using your own cohort usage curves instead of a flat contract."

"How do I model LTV for a usage-based product?"

"Chart monthly usage per account for recent cohorts, fit flat, median-growth and top-quartile scenarios, multiply by unit price and gross margin, and apply logo churn plus a soft-churn assumption for usage decline without cancellation. Compare to the flat-plan LTV for the same accounts."

"Is usage-based revenue harder to forecast?"

"Yes. Revenue follows customer behaviour. Commitments, prepaid credits, minimum fees and cohort-based forecasting restore predictability."

"What does usage-based billing infrastructure cost?"

"Hosted platforms often charge a percentage of billing volume; Stripe Billing lists pay-as-you-go at 0.7%. Self-hosted engines such as the ABAXUS core have no percentage fee but require engineering time to operate. ABAXUS Cloud is a fixed annual tier. Also price in ownership risk after the 2025 and 2026 acquisitions of OpenMeter, Metronome and Orb."

"When is a hybrid model better than pure usage-based pricing?"

"When usage growth is uncertain or customers want budget certainty. A base commitment with an included allowance gives you a revenue floor while metered usage above it keeps expansion automatic."

Usage-Based Pricing Economics: How CAC, LTV and NRR Change with Consumption Billing

Table of Contents

How Usage-Based Pricing Changes Unit Economics

Usage-based pricing changes more than the invoice. It changes how customer value accumulates, how revenue expands, and how well you can forecast. Standard SaaS metrics still apply, but they behave differently enough that they need recalibrating.

An earlier version of this post quoted benchmark ranges we could not trace to a source. This version uses figures from public filings and gives you a framework to model your own numbers, which is the only benchmark that matters for your pricing decision.

Net Revenue Retention: What Public Filings Show

Net revenue retention (NRR) is where usage-based models are strongest, and also where the data shows they are not magic.

Company	NRR figure	Source
Snowflake	158% (six months ended July 31, 2020)	S-1
Snowflake	125% (Q4 fiscal 2026)	Q4 FY2026 results
Twilio	140% (full year 2018), 128% (2017)	FY2018 10-K
Twilio	124% (Q4 2019) vs 147% (Q4 2018)	Q4 2019 results

Two lessons. First, usage-based companies can post very high NRR when customers are expanding workloads, as Snowflake did around its IPO. Second, the same mechanism works in reverse: when customers optimise spend or growth slows, usage-based revenue contracts without a cancellation. Snowflake’s rate fell from 158% to 125% with no change in its pricing model. Expect NRR to track customer activity, not just your product’s quality.

OpenView’s research on usage-based pricing also found higher net dollar retention among largely usage-based companies than others (TechCrunch summary); the direction is well supported, the exact levels depend on sample and year.

CAC and Payback

Usage-based pricing lowers the commitment needed to start, which changes acquisition in two ways:

Lower first-deal friction. A buyer can start with a small spend, often without procurement. Sales cycles for the initial deal can shorten, though we do not have a reliable general figure and you should measure your own.
Slower revenue ramp per account. The first invoice is smaller, so payback now depends on expansion. A subscription deal can pay back on contract value; a usage-based account pays back as usage grows.

Framework: model payback on a ramp, not a flat contract. If your blended CAC is $X per account, compute the month in which cumulative gross margin on a typical cohort passes $X, using your own usage growth curve.

LTV: Expansion Replaces Upsell

In a subscription model, lifetime value is mostly contract value times lifetime, with step-ups when you sell an upgrade. In a usage-based model, LTV is the integral of a usage curve that you do not control.

Framework for modelling LTV:

Take your last four quarterly cohorts and chart monthly usage per account.
Fit three scenarios: flat usage, median growth, top-quartile growth.
Multiply by your unit price and gross margin; apply your observed logo churn and a soft-churn assumption (usage decline without cancellation).
Compare against your flat-plan LTV for the same accounts.

If usage in your product does not grow with customer success, usage-based LTV will be lower than the flat plan. Check before changing pricing.

Forecasting

Usage-based revenue is harder to predict. Datadog’s filings note that it recognizes revenue for usage-based subscriptions as the product is consumed, which can cause results to fluctuate (10-Q). The levers that restore predictability are the same ones Snowflake-style businesses use: committed capacity contracts, prepaid credits and minimum commitments that create a revenue floor, plus cohort-based usage forecasts instead of account-by-account estimates.

Model it on real data

ABAXUS keeps usage events, aggregates and invoices in your own database, so cohort and unit-economics analysis runs on the same ledger you bill from. Self-host the free core, or use ABAXUS Cloud.

See how it works

The Cost Side: Billing Infrastructure

Usage-based pricing moves cost into infrastructure. You need exact metering, a pricing engine and an audit trail. Where that infrastructure comes from is itself an economic decision:

Percentage-of-volume fees. Stripe Billing’s pay-as-you-go rate is listed at 0.7% of billing volume. At $5M of monthly billing volume, that is $35,000 a month.
Fixed-fee or self-hosted options. A self-hosted engine such as the ABAXUS core has no percentage fee; you pay for the engineering time to run it. ABAXUS Cloud is the same engine hosted by us at a fixed annual tier.
Ownership risk. OpenMeter (Kong), Metronome (Stripe) and Orb (Adyen) all changed hands between September 2025 and mid-2026. Price in the cost of a forced migration.

See 7 Best Usage-Based Billing Platforms for SaaS in 2026 for the platform-by-platform comparison.

A Practical Model to Build

Input	Where to get it
Monthly usage per account, by cohort	Your metering data
Unit price and gross margin	Your pricing and cost accounting
Logo churn and soft-churn assumptions	Last 4 to 8 quarters of retention data
CAC by channel	Sales and marketing spend
Billing platform cost at your volume	Vendor quotes, or engineering cost if self-hosted

Output: payback month, 3-year LTV under three usage scenarios, and NRR sensitivity. If the usage-based case wins in the median scenario and survives the flat-usage scenario, you have a business case. If it only wins in the top-quartile scenario, run a hybrid model instead.

Conclusion

The economics of usage-based pricing are strongest where customer usage grows with customer success, and weakest where it does not. Public data shows both sides: very high NRR during expansion and sharp declines when customers optimise. Model your own cohorts, pick a plan shape that gives you a revenue floor, and make sure the metering behind the invoice is exact.