Usage-based billing charges customers for what they consume — API calls, gigabytes stored, tokens processed — instead of a flat fee. It powers cloud infrastructure and is spreading fast through SaaS, especially AI products. Three stages make it work: metering (counting usage), rating (turning counts into money), and invoicing (collecting without surprises).
Metering: how usage gets counted
Metering is the foundation — an engineering problem before it is a billing problem. The pipeline:
1. Define the billable metric — the single most important decision in metered pricing. Good metrics are valuable (customers understand why more costs more), measurable (reliably countable), and aligned (your costs scale with the same metric). API calls, gigabytes, messages, transcribed minutes, and tokens are classic examples. Bad metrics are things customers cannot predict or control — which is why "we meter everything" usually fails.
2. Emit usage events. As customers use the product, your application reports events to the billing system — via API, SDK, or streaming pipeline. Each event carries the customer identity, the metric, the quantity, and a timestamp. Modern platforms expose dedicated metering APIs for this (Stripe's Meters API, for example, stores events individually for audit).
3. Aggregate. Raw events are rarely billed directly — the billing system aggregates them per period using a rule you choose:
- Sum — total across the period (total API calls, total tokens); the most common.
- Maximum — peak value (peak concurrent seats, max storage). Useful when capacity, not consumption, drives cost.
- Last / latest — the most recent reading (seat count at period end).
- Average — mean across readings (average storage over the month, as cloud storage is typically billed).
- Count — number of distinct events or unique entities.
4. Handle edge cases. Idempotency (the same event reported twice must not bill twice), late-arriving events, and clock skew need defined behavior. This is where homegrown metering most often breaks: the billing math is easy, but the event pipeline's reliability determines invoice accuracy.
Rating: how counts become charges
Rating turns the aggregated usage number into a dollar amount. The common models:
Per-unit. Every unit costs the same: $0.01 per API call. Simple and transparent, but no volume incentive — largest customers pay the same rate as smallest.
Tiered (graduated). The unit price falls as usage crosses thresholds, with each tier's rate applying only to units in that tier — like income-tax brackets. The most common enterprise SaaS shape.
Volume (all-units). Cross a threshold and the entire usage reprices at the lower rate. Simpler to explain than tiered, but it creates cliff effects — use carefully.
Stairstep. Usage maps to fixed price bands. Predictable within a band, but customers near a boundary overpay or game the threshold.
Free allowance + overage. The first N units are included; usage beyond that bills per unit — the hybrid workhorse: a predictable floor with metered upside.
Minimum commitments with drawdown. The customer commits to $X per period and draws usage against it, paying overage beyond the commit. Common in enterprise contracts.
Most real products combine these: a base platform fee, an included allowance, tiered overage rates, and sometimes a commit. Rating is also where billing platforms differ most in capability — confirm yours supports the shapes you need before committing, because migrating rating logic later is painful (see how to migrate billing platforms).
Prepaid credits vs postpaid invoicing
Two fundamentally different ways to collect metered charges:
Postpaid (in-arrears). Usage accrues during the period and invoices at the end — the utility-bill model. Lowest friction for adoption (no upfront commitment), but the vendor carries collection risk and the customer gets a variable bill. Most B2B SaaS metered pricing works this way, usually with a base subscription underneath.
Prepaid credits. The customer buys credits upfront and burns them down with usage — the mobile-top-up model. Better cash flow and zero collection risk; the customer pre-commits spend. The complexity lives in the credit mechanics: expiry rules, unused credits, overage past the balance (hard stop vs. automatic top-up vs. postpaid overage invoice), and how credits interact with discounts.
Many products blend the two: a prepaid commit drawn down first, postpaid overage beyond it. AI products in particular have converged on credit-ish models because token costs are immediate and variable — see our usage-based billing for AI products guide.
The choice shapes your finance function: prepaid credits create deferred-revenue accounting, while postpaid creates receivables and dunning exposure. Involve finance before choosing.
Pure usage vs hybrid pricing
Pure usage-based (no base fee, pay only for consumption) has the lowest adoption friction and the most volatile revenue. It works when usage correlates tightly with value, starts near zero, and grows smoothly — cloud infrastructure, APIs, communications platforms. It fails when customers need budget predictability more than fairness, which in enterprise procurement is often.
Hybrid (base fee + metered usage) is the dominant SaaS pattern: the base fee guarantees minimum revenue and a predictable floor, while the metered component scales with value. One packaging lesson: start hybrid unless you have a specific reason to go pure.
The tooling landscape
Not every billing platform handles metering equally well — and the differences matter:
Stripe Billing handles straightforward metered pricing natively (meters, per-unit and tiered rating), but serious usage-based complexity historically went through Metronome — which Stripe acquired in January 2026 to close its usage-based gap, a roughly $1B deal deepening metered and hybrid billing (Baremetrics, Stripe vs. Recurly, September 2026). If your metering needs are growing on Stripe, that acquisition is the development to watch.
Purpose-built metering platforms — Orb and Metronome are the best-known — are usage-first billing engines built for high-volume event ingestion, complex rating, and real-time usage visibility. They fit AI and infrastructure products where metering is the billing core. Entry pricing is typically custom or quote-led at scale.
General billing platforms vary: Recurly and Chargebee both include usage-based billing; Maxio lists metering and rating among its capabilities; open-source options like Lago are usage-first and self-hostable. Enterprise suites (Zuora and similar) handle complex rating but with enterprise implementation timelines.
The build option. Some teams meter in-house and push rated amounts to their billing platform as invoice line items. This works at small scale, but the event pipeline, rating engine, and audit trail become your permanent maintenance burden. Buy the metering unless it is a core competency.
What makes metered billing hard
The mechanics are straightforward; the hard parts are operational:
Bill shock. A customer whose usage spikes 10x gets a 10x invoice. Mitigate with real-time usage dashboards, spend alerts and caps, and anomaly outreach before the invoice — treat "no surprises" as a feature.
Forecastability. Variable revenue is harder to forecast and variable bills harder to budget — hence minimum commits and hybrid packaging.
Metric design. The billable metric shapes customer behavior. Meter something customers cannot control and they will resent the bill; meter something disconnected from your costs and margins drift. Expect to iterate.
Revenue recognition. Metered recognition timing and prepaid-credit deferred revenue add accounting complexity — where a finance-grade billing platform earns its keep.
Disputes. "Your meter is wrong" is a harder support conversation than "I want to cancel." Invest in usage transparency — per-customer event logs, published aggregation rules — because invoices are only as trusted as the metering behind them.
When usage-based pricing wins
Usage-based pricing wins when:
- Value scales with consumption. The customer's benefit grows as they use more — API platforms, infrastructure, AI features, communications. If heavy users get proportionally more value, metering beats any seat count.
- Adoption friction must be near zero. No base fee means trying the product costs nothing — accelerating bottom-up adoption and letting small customers start small.
- Your costs are variable. When each unit costs you money (tokens, compute, bandwidth, per-message fees), metered pricing passes the cost structure through instead of averaging it into flat tiers — avoiding overcharged light users and margin-bleeding heavy ones.
- Expansion is the growth motion. In seat-based pricing, expansion means selling more seats; in usage-based pricing, it happens automatically as customers succeed. Net revenue retention tends to be higher — exactly why investors like it.
It loses when customers need budget certainty (much of enterprise procurement), when usage is spiky, when the billable metric is hard to understand or control, or when your costs are mostly fixed and metering adds complexity without aligning anything.
A final note: usage-based pricing changes the company, not just the pricing page — sales, finance, and product all shift toward consumption economics.
- Usage-based billing prices what customers consume (API calls, tokens, storage, minutes) via three stages: metering, rating, and invoicing.
- Metering aggregates events per period — sum, max, last, average, or count — and the event pipeline's reliability determines invoice accuracy.
- Rating models: per-unit, tiered (graduated), volume (all-units), stairstep, free allowance + overage, minimum commits with drawdown. Most real products combine several.
- Prepaid credits improve cash flow but create deferred-revenue accounting; postpaid lowers friction but creates collection risk.
- Hybrid (base fee + metered usage) dominates SaaS: predictability plus value alignment. Start hybrid unless you have a reason not to.
- Tooling: Stripe Billing covers simple metering natively (Metronome, acquired January 2026, is Stripe's usage-depth play); Orb/Metronome are usage-first engines; Recurly, Chargebee, Maxio, and Lago all handle metered models.
- The hard parts are bill shock, forecastability, metric design, revenue recognition, and disputes — invest in usage transparency from day one.
FAQs
What is usage-based billing?
Usage-based (metered, consumption) billing charges customers for what they actually use — API calls, gigabytes stored, tokens processed, minutes transcribed — rather than a flat subscription fee. Your product emits usage events; the billing system aggregates them per period, applies a rating model, and invoices the result.
What is the difference between metered and tiered pricing?
Metered describes how usage is counted (events aggregated per period); tiered describes how counts are priced (unit rates falling as usage crosses thresholds, each tier's rate applying only to units in that tier). Often combined: meter the usage, then rate it with tiered pricing.
Is usage-based pricing better than flat-rate?
Better when customer value scales with consumption, when near-zero adoption friction matters, when your costs are variable, and when automatic expansion drives growth. Flat-rate or hybrid is better when customers need budget predictability, usage is spiky, or the billable metric is hard to understand. Most SaaS products land on hybrid.
How do AI products bill for token usage?
Typically with a metered model: tokens (input/output, sometimes by model tier) are the billable metric, events stream to a metering system, and rating is per-unit or tiered — often wrapped in prepaid credits or a hybrid base-plus-usage plan. See our guide to usage-based billing for AI products.
Which billing platforms support usage-based pricing?
Stripe Billing (native metering; Metronome for depth), Orb and Metronome (usage-first), Recurly, Chargebee, Maxio (metered models included), Lago (open-source, usage-first), and enterprise suites like Zuora. Match the tool to your event volume and rating complexity.
Conclusion
Usage-based billing is conceptually simple — count what customers use, price it fairly, invoice it transparently — and operationally demanding in exactly the places flat-rate billing is not: the metering pipeline, the rating model, bill shock, and customer trust. Get the billable metric right, start with hybrid packaging, give customers real-time visibility into usage, and choose tooling that handles your actual event volume and rating shapes. Metered pricing aligns what customers pay with the value they receive better than any alternative — which is why it keeps spreading from infrastructure into mainstream SaaS. As with everything in billing, verify platform capabilities against your requirements before you commit; the pricing page is the smallest part of the decision.
—
Sources: Stripe's Metronome acquisition (January 2026, ~$1B reported) verified in Baremetrics' Stripe vs. Recurly (September 2026). Metering mechanics per standard industry patterns; confirm specific platform capabilities (ingestion scale, rating models, credit handling) on each vendor's current documentation before building.