What is usage rating?
Usage rating is the process of applying pricing rules to measured product consumption to determine its commercial value.
For an AI product, usage rating can convert token consumption, API requests, document processing, or AI workflow executions into credits, priced units, or monetary charges.
For example, a customer might consume 5,000 tokens while generating an AI response. The metering system records those tokens, while the rating system determines whether that usage should count as 10 credits, 50 credits, or a monetary amount.
Metering tells you how much was consumed. Rating determines how that consumption is priced.
This distinction is essential for AI products using usage-based billing.
How does usage rating work?
Usage rating operates between usage measurement and billing.
A simplified process looks like this:
Usage Event → Metering → Rating → Billable Usage
Consider an AI application that records the following event:
Customer: cus_123
Feature: document_summary
Provider: Provider A
Model: Model X
Input tokens: 4,200
Output tokens: 650
Total tokens: 4,850The metering layer captures the consumption.
The rating layer evaluates the applicable commercial rules.
For example:
Measured usage: 4,850 tokens
Applicable pricing: 50 credits per summary
Rated usage: 50 creditsThose credits may be deducted from the customer’s included allowance or handled under another billing arrangement.
The result depends on the product’s pricing model.
For a deeper explanation of the underlying event, see Usage Event.
Usage rating vs metering, costing, and billing
These concepts are closely connected, but they serve different purposes.
| Process | Question answered |
|---|---|
| Metering | How much was consumed? |
| Costing | What did that usage cost the company? |
| Rating | What commercial value should be assigned? |
| Billing | What does the customer owe? |
For example, an AI request might consume 4,850 tokens, cost the provider customer $0.036, and be rated as 50 credits.
If the customer has sufficient included credits, there may be no additional monetary charge.
Rated consumption does not automatically mean additional invoice revenue.
That distinction is explained further in our Billable Usage glossary.
Why usage rating matters for AI products
AI products often use multiple models, providers, and workflows with different cost structures.
A document summary and an autonomous research workflow may generate entirely different infrastructure costs.
Charging identical amounts for both could make pricing difficult to sustain.
Usage rating gives companies flexibility to define how consumption should count under their pricing model.
It helps teams:
- Apply different prices to different AI features or models.
- Translate technical usage into credits that customers can understand.
- Support customer-specific pricing for negotiated contracts.
- Adjust commercial rules without redesigning the underlying usage measurement.
- Protect margins by connecting pricing decisions to actual consumption costs.
For example, a basic model request might consume fewer credits than a premium model request, even if both process similar token volumes.
This is particularly useful for products using credit-based pricing.
Common usage rating models
AI companies can rate usage in several ways.
Fixed-unit rating: Each completed action receives a predetermined commercial value, such as 20 credits per document.
Volume-based rating: Usage is priced according to quantity, such as requests, tokens, or processing minutes.
Tiered rating: Different rates apply at different usage volumes.
Cost-based rating: The underlying infrastructure cost informs the commercial price, potentially with an additional margin.
Conditional rating: Pricing varies according to the model, provider, feature, customer, or other event attributes.
The right approach depends on how closely technical consumption reflects customer value.
Some products combine several models within one pricing system.
Common usage rating mistakes
One mistake is confusing provider cost with the customer price. A request costing $0.05 does not necessarily need to be sold for $0.05.
Another is applying the same rule to every model or feature without considering differences in consumption and cost.
Teams can also run into problems when pricing changes affect historical calculations. Billing systems should preserve enough information to explain how past usage was rated.
Finally, rating should not automatically turn every technical event into a customer charge. Internal requests, failed operations, free trials, and promotional usage may require different commercial treatment.
The wider separation between these stages is explained in How to Implement Usage-Based Billing for AI Services.
How MetricaOS supports usage rating
MetricaOS connects structured usage metering with configurable commercial pricing logic.
Teams can define usage metrics, associate activity with customers and models, configure provider pricing, and apply credit rules that determine how measured consumption affects customer balances.
This allows AI products to separate the technical measurement of usage from the commercial rules used to monetize it.
The resulting information can support customer wallets, subscription allowances, usage limits, billing, and cost analysis.
For AI companies, usage rating is an important bridge between product consumption and revenue.
Accurate metering establishes what happened. Flexible rating helps turn that usage into a sustainable pricing model.