Usage-Based Pricing: How to Choose the Metric and Structure It

Usage-based pricing means the customer pays for how much of your product they use: API calls, messages sent, documents processed, compute hours. Revenue grows as the customer grows, without a sales conversation to upgrade them.

Done well, it is the most natural expansion engine in software. Done badly, it produces bills nobody can forecast, finance teams that block renewals, and customers who ration their own usage. This guide covers how to choose the metric, how to structure it, and when not to use it.

When usage-based pricing works

It works when three things are true. Usage tracks the value the customer gets, so more use means more benefit. Your cost rises with usage, which is true of almost every AI product. And the customer can see and control their own usage. When all three hold, usage pricing aligns your revenue, your cost and the customer's success.

It works poorly when usage is a cost for the customer rather than a benefit. If more API calls mean your integration is inefficient, charging per call punishes customers for your product's waste.

Choose the metric first

The usage metric is the unit you charge for, and it matters more than the price per unit. Always look first for a metric close to the outcome the customer buys. Resolved tickets beat messages sent. Invoices processed beat API calls. When the outcome is hard to measure, choose the usage proxy that tracks it most closely.

We score each candidate metric with the ACE framework. Auditability: can the customer verify what they were charged for? Cost-Margin Fit: does revenue rise when your cost rises? Ease: can you bill it cleanly and can the customer forecast it? A metric that scores low on any one of the three will cause trouble at renewal.

Continuous or banded

Continuous pricing charges for every unit. It aligns closely with value but makes the bill hard to predict, and some buyers feel nickel-and-dimed by it.

Banded pricing groups usage into buckets, the way Dropbox sells storage plans. The customer picks a band and knows what they will pay. You give up some precision and gain predictability, lower friction and simpler procurement. For most B2B buyers, bands win.

Single or multiple metrics

One metric is the cleanest and easiest to understand. Use more than one only when there are genuinely separate sources of value, or when a cost unrelated to the core value has to be covered. Square charges a percentage plus a fixed fee per transaction for that reason: the fixed fee covers processing cost that does not scale with the payment's size.

The structures that work

Pay as you go. A price per unit with no commitment. Lowest barrier to start, highest bill volatility. Good for developer tools and self-serve products.

Prepaid commitment. The customer commits to a volume up front at a discount and draws it down. Predictable for both sides, and it gives you cash and a renewal moment.

Tiered volume. Bands of usage at falling unit prices. It rewards growth and is easy to forecast.

Hybrid: platform fee plus usage. A fixed fee covers access and a baseline volume, with usage charged above it. This is the default we recommend for AI products, because the fee covers your cost floor and the usage component captures growth. Our AI agent pricing guide shows a worked example.

Make the bill predictable

Buyers will pay more for a bill they can forecast. Give them the tools to do it. Show live usage in the product, send alerts at set thresholds, offer spend caps, and publish what a unit buys in plain language. If you sell credits, state the exchange rate: how many documents, calls or tasks a credit buys.

When not to use usage-based pricing

Avoid it when usage does not track value, when the customer cannot see or control usage, or when your buyers' budgets are fixed annually and any variance needs approval. In those cases a flat tier or a hybrid with a generous included volume serves both sides better.

Moving from seats to usage

Do not switch the whole base at once. Model what every existing account would pay under the new structure, find the accounts that would see large increases, and plan their migration with grandfathering or a phased ramp. Launch on new business first. Then test the price per unit with buyers using Van Westendorp and Gabor-Granger, both of which include a free calculator.

If your product does work rather than giving access to it, also read our outcome-based pricing guide. Usage is often the step between seats and outcomes.