All terms
Growth
Usage-based pricing
A pricing model where customers pay in proportion to what they consume (API calls, GB stored, seats active) rather than a flat subscription.
By Daniel Reyes · Last updated July 22, 2026
In plain English
You get paid when the customer gets value. Feels great when the customer scales; feels terrible when your customer has a quiet quarter.
Example
Snowflake: per-compute-second. OpenAI: per-token. Twilio: per-message. Vercel: per-build-minute + per-GB-bandwidth.
Why it matters
Usage-based aligns incentives beautifully (customer succeeds → you succeed) but destroys revenue predictability. Combined with a floor (minimum monthly commit) it becomes forecastable enough for boards.
Common mistakes
- Pure usage with no floor — one big customer's slow month tanks your ARR
- Metering the wrong unit (charging per API call when the customer's value comes from responses)
- Not showing usage transparently in the product — invoice-shock kills renewals