Usage-based billing is a pricing model where customers pay according to how much they consume of your product, rather than paying a fixed monthly price regardless of whether they use a lot or a little. If a customer consumes more, they pay more. If they consume less, they pay less. The price adjusts to the real value each customer receives.
Why this model is growing
Software has changed. AI products have real variable costs: every call to a language model, every generated image, every minute of video processing has a direct cost to the provider. Charging a flat fee when the cost of serving each customer varies drastically is not sustainable.
Usage-based billing aligns your revenue with your costs. Customers who get the most value from your product pay the most, and customers who are just starting out can try it without committing to a high price. This lowers the barrier to entry and increases retention: nobody cancels a subscription when they feel they are paying exactly for what they use.
Common usage-based billing models
There are three main models, and each one fits different types of products.
Metered
The metered model gives the customer an included amount of usage as part of their base plan and charges overages at the end of the billing period. For example, a $99/month plan includes 10,000 API calls. If the customer makes 15,000 calls, they are charged $0.01 for each of the 5,000 additional calls when the cycle closes.
This model works well for products where usage is predictable and gradual, such as APIs, email platforms, or storage services.
Credits
The credits model assigns the customer a number of credits with their plan. Each action within the product consumes credits. When credits run out, the customer can purchase additional credit packs or wait for the next billing cycle.
It is the preferred model for generative AI products. ChatGPT, Midjourney, and similar tools use credits because each generation has a variable cost, and the credits model makes consumption tangible for the user.
Balance
The balance model converts the plan's base price into a real-money spending balance. Each customer action deducts an amount from that balance. If the balance runs out, overage is charged at the end of the period.
This model is ideal for cloud-style infrastructure platforms where customers need exact visibility into how much they are spending in dollars, not abstract units.
When to use usage-based billing
Usage-based billing makes the most sense when the cost of serving your customers varies significantly between them. If your product has high marginal costs per active user, such as AI processing, bandwidth, or compute, a usage-based model protects your margins.
It is also the best choice when you want pricing to be an acquisition tool. A low or zero entry price, where the customer only pays when they consume, eliminates the price objection during evaluation.
It is less ideal if your product delivers the same value regardless of how much it is used, such as project management tools or CRMs. In those cases, a fixed seat-based pricing model is often more appropriate.
Examples in production
The most successful products of recent years use usage-based billing: AI tokens at OpenAI and Anthropic, API calls at Twilio and SendGrid, storage and compute at AWS, video minutes at transcription platforms. The pattern is clear: when the cost of producing value is variable, the price should be variable.
How Commet implements usage-based billing
Commet supports all three consumption models as a native part of its plan system. Each plan can be configured with a metered, credits, or balance model. Usage event ingestion happens through the SDK, and Commet automatically calculates overages, manages billing cycles, and generates the corresponding invoices at the close of each period. For a practical walkthrough, see the guide on usage-based billing for AI products.