Usage based pricing: the models that work, with examples

Usage based billing sounds like one idea. It is five distinct patterns, plus a variant worth knowing, and picking the wrong one for your product creates problems that show up months later as billing disputes and churn. The patterns below are the ones that survive contact with real products, with the tradeoffs of each.

1. Pure per unit

You pick a unit and a price, and the bill is the multiplication. A messaging API might charge $0.005 per message sent. An email service might charge $1 per 1,000 deliveries.

This is the easiest model to explain and the hardest to predict. Customers love the fairness and hate the surprise bills. It works best when usage maps directly to value the customer can feel, and when a single month of usage is not wildly different from the last.

2. Tiered pricing

The price per unit changes as usage climbs, and each band is charged at its own rate. For example: the first 1,000 API calls at $0.10 each, the next 9,000 at $0.05, everything beyond at $0.02.

A customer who uses 12,000 calls pays 1,000 × $0.10 plus 9,000 × $0.05 plus 2,000 × $0.02, which is $590. The early units always cost more. This rewards growth without giving away the first units cheap.

3. Volume pricing

The whole quantity is charged at the rate of the band it lands in. The same 12,000 calls under volume pricing would all be charged at the $0.02 rate: $240. Volume pricing is simpler to reason about and cheaper for big customers, but there is a quirk to watch: crossing a band boundary can make a bigger usage month cost less than a smaller one right below the boundary. Some teams use stepped pricing instead, where the band sets one flat price for the whole quantity, so the invoice is a single predictable number and the boundary quirk disappears.

4. Included allowance plus overage

A monthly subscription includes a bundle of usage, and only usage beyond it is metered. A $99 plan might include 100,000 events, with overage at $0.50 per 1,000 after that.

This is the most common pattern in subscription products for a reason. You get predictable base revenue. The customer gets a predictable floor. Heavy users pay more without anyone renegotiating. The design work is sizing the allowance so the plan price feels honest.

5. Prepaid credits

Customers buy a balance up front and usage draws it down. Common in AI products, where a customer might buy 10,000 credits for $90 and each request burns credits based on the work done. We wrote a separate guide on when credits beat subscriptions, and the token pricing guide covers the AI specific mechanics.

Picking between them

A few questions cut through most of the debate:

• Can the customer predict their own usage? If yes, pure per unit is fine. If no, wrap it in an allowance or credits so the bill has a floor and a ceiling they chose.
• Does your cost scale with their usage? If your costs are mostly fixed, aggressive tiering is a gift you can afford. If costs scale linearly, your bands need to protect margin at every size.
• Who reads the invoice? If a finance team approves it, they will ask how the number was calculated. Pick a model you can explain on the invoice itself, not one that needs a meeting.

The API pricing guide walks these questions for API products specifically.

Whatever you pick, show the math

The single biggest source of usage billing disputes is not the price. It is that the customer cannot see how the number on the invoice came from the usage they remember. If your invoice says $590, it should be able to show the bands, the counts in each, and the per band subtotals. Mid cycle plan changes have their own version of this problem, which is why we wrote up proration with worked numbers.

This is the part we built yRecurring around. All of the models above are built in (flat, per unit, tiered, volume, stepped, package, percentage, and dynamic), they can be combined on one subscription, and every amount shows how it was worked out: the formula, the inputs, every step. The page on pricing models walks through each one, and usage and credit billing for AI shows how they fit together for AI and API products. If you want to model your own pricing and read the invoices it produces, the trial is free and takes minutes to set up.