Credits-based billing is a pricing model where customers receive a fixed number of credits as part of their subscription. Every action they take in the product consumes one or more credits. When credits run out, the customer either purchases additional credit packs or waits until the next billing period resets their balance. Unlike metered billing, credits create a hard spending limit that prevents surprise charges.
How credits-based billing works
A credits plan has three parts: a base price, an included credit amount that resets each billing period, and a per-credit price for purchased credit packs.
Consider an AI writing tool at $49/month that includes 500 credits. Each document generation costs 10 credits. The customer can generate 50 documents per month. On day 18, they have used 480 credits. They can generate two more documents before they hit the wall. At that point, they either buy a credit pack (say 200 credits for $19) or wait until the billing cycle resets their balance back to 500.
This flow is familiar to anyone who has used ChatGPT Plus or Midjourney. You get an allocation, you use it, and when it runs out, you decide whether to buy more.
Plan credits vs. purchased credit packs
There is a critical distinction between the two types of credits a customer can hold.
Plan credits come included with the subscription and reset at the start of each billing period. If a customer on a $49/month plan has 120 unused credits when the cycle ends, those credits disappear. The new period starts fresh at 500. Plan credits are not rollover currency.
Purchased credit packs are bought separately and never expire. If a customer buys a 200-credit pack on March 5 and only uses 60 of those credits before the March billing period ends, the remaining 140 carry forward into April, May, and beyond. They are the customer's property until consumed.
The system always consumes plan credits first. Purchased credits are only touched after plan credits are fully depleted. This protects the customer's purchased credits and incentivizes them to use their subscription allocation efficiently.
For setup details on credit packs, see the documentation on credit packs.
How credits differ from metered and balance billing
Credits, metered, and balance billing are three distinct consumption models, and a plan uses exactly one of them.
With metered billing, the customer is never blocked. They exceed their included quota and pay the overage at the end of the period. This makes metered billing seamless but unpredictable for the customer. A usage spike can produce a bill that is three or four times the expected amount.
With balance billing, the plan's base price becomes a dollar balance. Usage deducts real monetary amounts. The customer thinks in dollars, not abstract units.
Credits sit in the middle. They create a predictable cost ceiling (the customer knows exactly what they will spend this month), and they give the product team control over when to upsell. The moment a customer runs out of credits is a natural conversion point for a credit pack purchase or a plan upgrade.
When to use credits-based billing
Credits work best when individual actions have variable cost to you but you want the customer to think in simple, whole-number units rather than fractional dollars.
AI products are the clearest example. A single image generation on Midjourney might cost the provider anywhere from $0.002 to $0.05 depending on the model and resolution. Exposing that price variation directly would confuse users. Credits abstract it: "one standard image = 1 credit, one high-res image = 4 credits." The customer understands the trade-off without needing to think about infrastructure costs.
Credits also work well for marketplace-style platforms where different actions have different costs. A design tool might charge 1 credit per template export and 5 credits per custom design render. The credit becomes the universal currency across the product.
Credits are a poor fit when usage is continuous and the customer cannot meaningfully control their consumption. A monitoring tool that ingests all incoming logs automatically would frustrate users who cannot stop credit consumption without turning the product off.
Real-world credit systems
ChatGPT Plus gives subscribers a monthly message allocation for GPT-4. Once you hit the limit, you drop to the base model or wait. Midjourney gives subscribers GPU minutes as credits, with different generation types consuming different amounts. Canva Pro includes credits for AI image generation that reset monthly.
The pattern is consistent: a generous base allocation that covers the typical user, a clear exhaustion point, and an upsell path (buy more or upgrade).
Implementation considerations
Credit systems need a real-time balance check before every action. If the balance check happens asynchronously and the customer fires off five requests simultaneously, you can end up with a negative balance. This requires either optimistic locking with a post-check or a synchronous balance reservation before processing.
Related
- Metered Billing: usage tracked, overage charged at period end, customer never blocked
- Balance Billing: plan base becomes a dollar balance, usage deducts real money
- Usage-Based Billing: overview of all three consumption models
- Consumption Models: how to configure metered, credits, or balance in your plan
- Credit Packs: setting up purchasable credit packs for your customers