Usage-based pricing simply means customers pay for what they use. It has become the default for modern software because value and activity are uneven, especially in AI. Seat-based pricing made sense when headcount drove adoption and marginal costs were low. Today, costs and value often track backend activity, not user count. That shift demands pricing that scales with usage and an architecture that lets you change pricing without weeks of engineering work.
Usage can mean many things: API calls, tokens, minutes, gigabytes, events, or actions.
Twilio: Pay per SMS message sent (and, in many cases, received), with per-country rates so spend scales directly with message volume.
Postmark: Charge per email sent, often with included allowances and straightforward overage pricing as volume grows.
AWS: Metered by storage (GB-month), requests, data transfer, and compute milliseconds—your bill reflects exactly the resources you consume.
Vercel: Serverless function invocations and bandwidth drive spend; more traffic or heavier compute means more usage, no manual capacity planning.
OpenAI / Anthropic: Token-based billing where inputs and outputs are priced per token; larger models and longer prompts/responses consume more.
Synthesia: Video generation priced by minutes/credits; longer renders and higher resolution consume more credits.
Utilities: Electricity by kWh and water by gallons; simple pay-for-what-you-use models, sometimes with a small base fee.
Uber: Fares based on time and distance, with demand-based adjustments; usage (the trip) directly drives price.
Different shapes exist because teams trade off speed, predictability, discounting at scale, and budget control. The four most common models below cover most needs.
Pay-as-you-go (PAYG)
The on-ramp to usage-based pricing. You pick a unit, set a public rate, meter usage each period, and invoice rate × usage. It’s fast to ship, easy to explain, and great when one or a few meters map cleanly to value. Read more: Pay-As-You-Go 101 and PAYG vs. Alternatives.
Learn more: PAYG introduction
Overage pricing
Each plan includes an allotment; extra usage bills at a posted overage rate. Buyers get a predictable baseline; you capture upside as usage grows. Simple to implement and familiar to customers. Read more: Overage Pricing 101 and How Overages Compare to Other Models.
Learn more: Overage pricing introduction
Tiered (graduated/volume)
Still usage-based, but unit cost falls as usage grows. It rewards scale, calms spend anxiety, and mirrors enterprise expectations without bespoke deals. Read more: Tiered Pricing 101 and The Tiered Pricing Playbook.
Learn more: Tiered pricing introduction
Credit burndown
Customers pre-buy a pool of credits and draw down as they perform actions. It hides technical inputs and shows buyers one balance across features, which improves budget control and gives you flexibility to adjust burn without repackaging plans. Read more: Credit Burndown for Fast-Growing Companies and Credit Burndown vs. PAYG.
Learn more: Credit Burndown
Usage-based models span a spectrum, from fast, minimal setups to more structured systems with extra controls. Simpler models ship quickly and are easy to explain, while richer models add predictability, discounting, and governance at the cost of complexity.
Figure out what buyers care about most (e.g. cost certainty, easy visibility into usage, and rewards for higher volume). Meter the behaviors that signal value, ship the simplest pricing that fits, validate it with customers, then layer in complexity as real needs show up.
Usage-based pricing aligns revenue to real customer activity. Start with the simplest model that matches how value shows up in your product, then iterate as you learn. If you want to ship these models without hard-coding pricing logic, Schematic gives you metering, entitlements, credit ledgers, and plan iteration so pricing changes are a day-scale task, not a quarter-scale project.
