What is Token-Based Pricing?

Token-based pricing charges customers by standardized units of consumption called tokens. In SaaS and AI products, a token represents a measurable unit of work — text processed by a Large Language Model, sent to it or generated by it, or an API call. Unlike seats or flat subscriptions, cost tracks actual consumption. 

A token is a unit representing a particular type of activity on a platform; the token, being a unit of work, an abstract standard, is a completely different concept from a ​‍​‌‍​‍‌seat/license.

Token pricing is generally given as a rate per 1,000 or 1,000,000 tokens separately for input and output. Companies arrange things in several ways:

• Pass-through: model-provider costs billed on at cost, no markup.
• Marked-up rates: They are determined by adding a margin to the provider’s costs.
• Credit packages: tokens bundled into prepaid credits.
• Hybride: the combination of the token usage charge and base subscription.

To reduce exposure to potential customer overconsumption, token systems offer prepayment credit wallet facilities with auto-approach alerts and planned surcharge pricing. As customers near their credit balance threshold, alerts are received; after credits are expended, subsequent usage generally aligns with one of three patterns:

• Cessation of activity is conditional on credit replenishment
• Subsequent usage is subject to the overage rate determined earlier

All usage data needs to be tracked to get a proper understanding of how the token system is going. After having event-traceable logs, you will want to ensure aggregations are accurate and the rating logic used is consistent. Then, construct traceability that makes the billing process defensible:

• Pricing rules that allow for reversion, and documentation that explains procedures
• Regular checks and audits of the accuracy of the “meter-to-invoice” relationship are a regular task of the billing department

Token pricing changes how revenue is recognized under standards like ASC 606. Revenue is recognized as tokens are consumed, not when the customer pays. A prepaid pack sits as deferred revenue and draws down incrementally: buy 1,000 credits, burn 300, and those 300 are revenue this period — not when the balance hits zero.

Breakage is the practice of accounting for prepaid tokens that were purchased but not redeemed by their owner before expiry. Estimating and recognizing breakage are two separate processes. From a company’s perspective, breakage can be recognized as revenue where historical data suggests customers will not redeem their unused credit.

Token pricing introduces factors that require careful evaluation concerning customer sentiment and financial outcomes:

• Variations in usage can correspond to differences in billing totals.
• Revenue and cost planning involves complexity.
• The relationship between token costing and selling price affects the resulting margin.
• Compared to the existing standard, metering infrastructure implementation requirements and considerations for compliance risk are positioned at an elevated level.