Glossary
Usage-based pricing
The bill follows consumption, which cuts both ways for buyer and vendor.
Definition
Usage-based pricing is a model in which the amount a customer pays is determined by a consumption metric such as API calls, gigabytes stored, messages sent, tasks completed or hours delivered, rather than by a fixed subscription fee. Charges are normally metered continuously and billed in arrears, so a customer who uses nothing in a period pays nothing or only a small platform fee.
- Also called
- Consumption-based, pay-per-use
- Origin
- Cloud infrastructure, mid-2000s
- Billed
- In arrears, on a meter
Where it came from and why it spread
Metered billing is old, since utilities have charged this way for a century, but its arrival in software came with cloud infrastructure in the mid-2000s, when Amazon Web Services made per-hour compute and per-gigabyte storage normal. Developer tools followed, then communications and payments, then everything with a countable unit.
OpenView's State of Usage-Based Pricing survey found adoption among SaaS companies rising sharply through the early 2020s, reporting that around 45% of surveyed companies had adopted some form of usage-based model by 2021. Later surveys disagree on the exact figure, partly because hybrid models, a platform fee plus a meter, blur the definition.
The variants, which are not interchangeable
Pure pay-per-use
No commitment, no floor. Pay for what the meter records. Rare in enterprise software because the vendor's revenue becomes hard to forecast.
Hybrid
A platform or seat fee plus a meter for consumption above an included allowance. The most common shape in practice.
Prepaid credits
Buy a balance, draw it down. Popular with AI products because it caps the customer's exposure and improves the vendor's cash position. Unused credits often expire, which is where trust is lost.
Outcome-based
Charging per completed unit of work rather than per resource consumed: per resolved ticket, per delivered hour. Aligns price with value and requires a definition of done both sides accept.
The honest trade-offs
What buyers gain and lose
- No cost for users or capacity that go unused
- Cost scales with value received rather than headcount
- Budgeting is harder: the bill varies month to month
- Fear of runaway usage discourages exploration unless caps exist
What vendors gain and lose
- Low barrier to starting, since small usage costs little
- Revenue expands automatically as customers grow
- Forecasting is harder and churn is quieter, since usage drops before anyone cancels
- The meter must be legible, or every invoice becomes a support ticket
Commonly confused with
| Term | What it means | The difference |
|---|---|---|
| Pay-as-you-go seats | Monthly rather than annual per-user billing | Still per seat. Only the commitment length changed, not the unit. |
| Freemium | A free tier with paid upgrades | A packaging decision about who pays, not a decision about what the meter counts. |
| Tiered pricing | Fixed price bands with feature or volume limits | The price is a step function chosen in advance rather than a continuous meter. |
| Overage charges | Fees for exceeding a plan's allowance | A meter bolted onto a subscription, usually priced punitively rather than proportionally. |
Related terms
Polaris uses the outcome-based variant with no platform fee: the software is free, and billing starts only when an AI worker delivers work, at roughly two dollars per human-equivalent hour.
Per-seat pricing
The bill tracks headcount, which is only a proxy for value received.
Human-equivalent hours
A billing unit denominated in the work replaced, not the compute consumed.
Work log
The record that makes an unwatched run reviewable afterwards.
Per-seat pricing and usage pricing charge for different things
One model prices access. The other prices output. Almost every argument about software cost is really about which of those you are buying.
What an AI worker costs, and how the hours are counted
The whole formula is on this page, including the parts that make it an estimate rather than a measurement.
What your work stack actually costs
Sixteen pages of pricing arithmetic you can check yourself, with the billing basis named on every row.
Questions people ask
+Is usage-based pricing cheaper than a subscription?
It depends entirely on usage relative to the subscription's break-even point. Light and uneven users almost always pay less; heavy consistent users can pay more. The reliable difference is not the total but the shape: cost tracks activity instead of headcount.
+How do buyers protect themselves from a surprise bill?
Spending caps, alerts at defined thresholds, and a meter they can inspect in near real time. A vendor that cannot show consumption as it accrues is asking for trust that the invoice will not justify at the end of the month.
+Why do AI products favour it?
Because their marginal costs are real. Every inference costs the vendor money, so a flat subscription either overcharges light users or loses money on heavy ones. Metering passes the underlying cost structure through instead of averaging it across everyone.
+What makes a usage meter trustworthy?
Three properties: the unit is something the buyer recognises as valuable, the calculation is published rather than proprietary, and each charge is itemised against the activity that produced it so it can be disputed individually.
Related
Per-seat pricing
The bill tracks headcount, which is only a proxy for value received.
Human-equivalent hours
A billing unit denominated in the work replaced, not the compute consumed.
Work log
The record that makes an unwatched run reviewable afterwards.
Tool sprawl
The expensive part is not the licences. It is that nothing is authoritative any more.
Per-seat pricing and usage pricing charge for different things
One model prices access. The other prices output. Almost every argument about software cost is really about which of those you are buying.
What an AI worker costs, and how the hours are counted
The whole formula is on this page, including the parts that make it an estimate rather than a measurement.
The cost of AI subscriptions for a team
Five figures a year at fifty people, and no shared record of what any of it produced.