Markets and business models
Usage-based pricing
Customers pay for what they consume. Spend tracks value — and your costs often track usage too.
Usage-based businesses charge for what the customer consumes: API calls, tokens, gigabytes scanned, messages sent, compute credits. The pitch is alignment — customers pay when they get value, and vendors grow as customers succeed. The catch is that your costs often scale with usage too. That is a different animal from classic seat-based SaaS, where delivering one more login is nearly free.
Twilio (messages), Snowflake (credits), Stripe (payment volume with a usage feel), AWS and model APIs all live here. AI application companies that resell foundation-model tokens live here whether they admit it or not.
Concept
COGS are the direct costs of delivering what you sold — hosting, inference, carrier fees, payment processing. Gross margin = (Revenue − COGS) / Revenue. Seat SaaS often runs ~70–85% at scale; infrastructure and AI APIs frequently land closer to 50–60% or lower early on. See gross margin.
Why usage revenue is seductive
Alignment with value. A startup sending 500 SMS verifications and an enterprise sending millions both start on the same API. Spend tracks use. That lowers adoption friction versus a $50k/year seat contract before the first call.
Land and expand without seats. Expansion is not "buy more licenses" — it is "traffic grew." Healthy usage businesses can post strong net revenue retention when successful customers consume more.
Developer-led adoption. A credit card and an API key beat a six-month sales cycle for the first wedge.
The downsides are real:
- Revenue volatility. Customers can cut usage overnight. Seat SaaS churns logos; usage businesses churn dollars inside logos.
- Forecasting pain. Finance hates "it depends on their traffic." You need commits, minimums or hybrid packaging.
- Margin discipline. If COGS are 40–50% of revenue, you have less room for S&M mistakes than an 80% margin SaaS peer.
Trade-off
Pure usage vs seat vs hybrid. Pure usage maximizes alignment and expansion but maximizes volatility. Pure seats maximize predictability but cap expansion. Hybrid (platform fee + included usage + overage, or commit + on-demand) is the usual mature form.
Walking through a hybrid
A $20/month flat plan with "unlimited" tokens looks simple. If a user consumes 5M tokens and upstream cost is $8 per 1M plus $2 other COGS, that user costs you $50 against $20 of revenue — negative 150% margin. A credit pack ($20 includes 1M; overage at $25 / 1M) keeps light users healthy and prices the whale. Full arithmetic is on AI wrapper economics.
Common mistakes
- Charging seats when value is API calls. Light users feel punished; heavy users are undercharged.
- Unlimited plans plus variable COGS. Your best customers become your worst.
- Assuming usage NRR behaves like seat NRR. The same logo can cut 40% of spend without "churning" in the CRM.
Check yourself
Why is usage revenue more volatile than seat SaaS?
Customers can cut consumption overnight without cancelling. Dollars leave inside a logo that still looks retained.
When does hybrid packaging beat pure pay-as-you-go?
When you need forecastability (commits, platform fees) without giving up alignment on overage.
Key takeaways
- Usage pricing aligns spend with value and enables developer-led land.
- Revenue is volatile and COGS scale with success.
- Hybrid (base + included usage + overage, or commit + on-demand) is the mature form.
- Always model power users. Unlimited plus variable COGS loses money on whales.
Further learning
Next, why infrastructure businesses have real COGS.