Skip to main content

AI Billing Definitions

Canonical definitions for margin analytics, cost attribution, and unit economics terminology used across Bear Billing.

Last updated: December 2025

Customer Margin

Core Concepts

Customer Margin is revenue minus variable costs attributable to a specific customer. Shows profitability before fixed costs.

Formula

Customer Margin = Revenue - Variable Costs

Example

If a customer pays $500/month and their AI infrastructure costs $150/month, their margin is $350 (70%).

Cost-to-Serve

Core Concepts

Cost-to-Serve is the total variable cost of delivering your product or service to one customer, including API costs, compute, and per-customer overhead.

Formula

Cost-to-Serve = (API Costs + Compute + Per-Customer Overhead) / Customer

Example

A customer using 500K tokens/month at $0.03/1K tokens has API cost of $15. Add $5 compute overhead = $20 cost-to-serve.

Usage-Based Billing

Core Concepts

Usage-Based Billing is a pricing model where charges are based on actual consumption (tokens, API calls, compute time) rather than flat subscription fees.

Example

Charging $0.10 per 1,000 API calls instead of a flat $50/month subscription. Revenue scales with customer usage.

Profit per Customer

Core Concepts

Profit per Customer is the direct revenues and costs associated with a single customer, used to assess business viability.

Formula

Profit per Customer = LTV - CAC - Cost-to-Serve

Example

Customer LTV of $2,000, CAC of $500, and lifetime cost-to-serve of $800 = $700 profit per customer.

Margin Health

Assessment Terms

Margin Health is a qualitative assessment of customer profitability, typically categorized as Healthy (>40% margin), At Risk (20-40%), or Unhealthy (<20%).

Example

A customer with 15% margin is "Unhealthy" and should be repriced or upgraded to a higher tier.

Token Cost Attribution

AI-Specific Terms

Token Cost Attribution is allocating the cost of LLM API tokens to specific customers based on their actual usage.

Formula

Token Cost = Tokens Used × Cost per Token

Example

Customer uses 1M input tokens at $3/1M and 500K output tokens at $15/1M = $3 + $7.50 = $10.50 total token cost.

Model Economics

AI-Specific Terms

Model Economics is the revenue and cost breakdown for a specific AI model (e.g., GPT-4o vs Claude Sonnet), showing margin per model.

Formula

Model Margin = Model Revenue - Model Costs

Example

GPT-4o generates $10K revenue but costs $4K = 60% margin. Claude Sonnet generates $8K revenue at $2K cost = 75% margin.

Power User Problem

Common Problems

Power User Problem is when heavy users on flat-rate pricing consume disproportionate resources, creating negative margins despite appearing profitable.

Example

A $99/month customer using $300/month in API calls has -200% margin. They appear profitable in revenue reports, but the variable cost exceeds revenue.

Revenue Leakage

Common Problems

Revenue Leakage is uncaptured or undercharged revenue due to billing gaps, reconciliation errors, or pricing model misalignment.

Example

Using a flat rate when usage varies substantially between customers means high-usage customers receive value not captured in pricing.

Pricing Scenario

Modeling & Simulation

Pricing Scenario is a simulated pricing model applied to historical usage data to project revenue and margin impact before deployment.

Example

Testing a tiered model against current flat-rate shows 15% revenue increase and 20% margin improvement before changing prices.

See These Concepts in Action

Use our calculators to apply these definitions to your own data.