Usage-based pricing isn’t just a pricing model—it’s a strategic framework that requires four core decisions: what to meter, which unit to charge, how to structure rates, and how to manage commitments and overages. Many teams discover too late that a flawed design can lead to disputes, churn, or margin erosion. The key isn’t just adopting usage-based pricing; it’s designing it with precision.
Understanding usage-based pricing in practice
Usage-based pricing (UBP) flips the traditional SaaS model by charging customers for what they actually consume, rather than a fixed subscription fee. The most effective UBP models tie billing directly to value delivered or resources consumed. Common examples include:
- API calls: Twilio charges per delivered message, aligning costs with successful outcomes rather than raw requests.
- Compute time: Snowflake bills by compute-seconds, rewarding efficient queries and penalizing inefficient ones.
- Data volume: Datadog charges for gigabytes of data processed, ensuring costs scale with usage.
- Output size: DeepL bills per translated character, directly linking cost to delivered value.
The model can take different forms—pay-as-you-go, prepaid credits, or committed minimums with overages—each suited to different business goals. The challenge lies in selecting the right combination of meter, unit, rate structure, and contract terms.
Selecting the right meter: the foundation of fair billing
The meter determines what you count, and a poor choice can create long-term friction. A strong meter must satisfy four critical criteria:
- Correlate with customer value: If a customer’s business growth relies on successful transactions, charging for successful outcomes (like Adyen does) ensures their bill rises only when they succeed. Charging for attempts—regardless of outcome—feels like a tax and erodes trust.
- Align with your cost structure: Your gross margin depends on matching billing to your cost of goods sold (COGS). For an AI company, charging per token consumed (as with inference costs) is logical. A flat per-request model could devastate margins if customers send unusually long prompts.
- Ensure auditability: Both you and your customer must independently verify usage. If raw event logs can’t reconstruct yesterday’s bill, disputes will arise. Transparency builds trust.
- Maintain stability: Frequent redefinitions of metrics—like changing what constitutes an "active user" quarterly—disrupt customer forecasting and renewal processes.
Most B2B products meter either events (API calls, document generations, workflow runs) or resources over time (compute-seconds, storage-gigabytes-per-month, active users). Mixing both complicates billing and confuses customers. Stick to one.
Example: Twilio could have charged per API call, but instead it bills per delivered message. A customer sending 10,000 messages that only 9,000 deliver pays for 9,000—fairly reflecting the service’s value and Twilio’s responsibility. Snowflake’s choice to bill by compute-seconds, not queries, ensures inefficient queries cost more, aligning customer behavior with operational costs.
Crafting the rate card: simple structures with big impact
The rate card defines the price per meter unit, and its design directly influences adoption, retention, and profitability. The three primary structures are:
Linear pricing
Linear models charge a fixed rate per unit, regardless of volume. For example, $0.01 per API call. This works best when:
- Your COGS are linear (each unit costs the same to deliver).
- Competitors use similar models, making differentiation difficult.
- Your customer base has consistent usage patterns.
Linear pricing is easy to explain but can limit scalability if costs rise disproportionately at high volumes.
Tiered pricing
Tiered models adjust rates at predefined thresholds. For instance:
- First 100,000 calls: free
- Next 1,000,000 calls: $0.02 each
- Above 1,100,000 calls: $0.005 each
This structure accommodates heterogeneous usage patterns—ideal for products serving both small startups and large enterprises. Vercel, Datadog, and AWS use tiered models to balance accessibility with margin protection.
Volume discounts
Volume discounts apply a single per-unit rate once a threshold is crossed, simplifying billing but complicating internal modeling. For example:
- All usage above 1,000,000 calls billed at $0.01 per call, regardless of volume.
This approach is easier for customers to understand but requires careful financial planning to avoid margin erosion.
Avoid penny pricing: Charging $0.0001 per token might seem precise, but it obscures real costs for customers. Rounding to $0.01 or bundling units (e.g., per 1,000 tokens) improves clarity and trust.
Managing commits and overages: balancing predictability and flexibility
Most companies start with pay-as-you-go pricing but transition to committed minimums as customer relationships mature. A well-structured commit system typically includes:
- Base commit: A minimum annual or monthly spend the customer guarantees.
- Overage billing: Charges for usage beyond the commit at the agreed rate.
- Flexibility clauses: Options to adjust commits based on seasonal or growth-related changes.
The goal is to provide predictability for your finance team while allowing customers to scale without fear of unexpected bills. Overly rigid commits can stifle growth, while overly loose ones reduce revenue visibility.
As usage-based pricing evolves, teams must prioritize transparency, fairness, and alignment with both customer value and business sustainability. The right design doesn’t just improve billing—it strengthens customer relationships and drives long-term growth.
AI summary
Learn how to design usage-based pricing that aligns with customer value and your costs. Avoid costly redesigns by making these 4 critical decisions early.
