Usage-Based Pricing: Value Metrics, Free Allowances and Real Examples

In usage-based pricing, the price is a rate multiplied by a quantity. Datadog charges $15 per host per month on Infrastructure Pro. Algolia charges $0.50 per 1K search requests on Grow. GoCardless takes 1% plus £0.20 per transaction on Standard.

Pure usage pricing has no plan grid at all. Groq lists model prices per million tokens. Replicate charges by the second for hardware time. The pricing page is a rate card.

Usage pricing fits best when customer value varies a lot from account to account. A side project and a large production app can both run on Algolia. A single flat plan would overcharge the first and undercharge the second. Metering lets both pay a price that matches what they get.

Most companies do not run pure usage pricing. They mix it with a subscription, credits or tiers. A base fee plus metered usage is hybrid pricing, and prepaid units spent on actions is credit-based pricing. Our SaaS pricing models guide compares all seven models side by side.

The value metric is the unit you charge for. Get it right and revenue grows as customers succeed. Get it wrong and customers feel taxed for using your product.

Look at how the catalog handles it. GoCardless charges a percentage plus a fixed fee per transaction, so its revenue moves with the money its customers collect. EasyPost charges per shipment and takes 1% of shipment value for insurance. Groww charges ₹20 per executed order with ₹0 account opening and maintenance. In each case the unit is something the customer already counts in their own business.

Compare that with compute units, resource hours or abstract tokens. They track the vendor's costs well but mean nothing to most buyers until translated. If your metric needs a glossary, plan to do that translation on the pricing page.

A good usage metric passes four tests:

• It tracks customer value. More of the unit means the customer got more out of the product.
• It is easy to understand. Buyers can estimate it without reading docs. Shipments and transactions pass. Compute units usually need explaining.
• It is predictable. Customers can forecast next month within a reasonable range.
• It tracks your costs. If serving the unit costs you money, charging for it protects your margin.

Pure pay-as-you-go has the lowest barrier to entry. Blaxel's pay-as-you-go plan starts at $0 plus usage, and its hero states the model twice so no one misses it. The tradeoff is revenue that moves with every customer's traffic.

A base fee fixes that. Langfuse charges $29 per month on Core, plus $8 per 100k units beyond the included amount. Helicone pairs a base fee with metered usage and backs it with a live pricing calculator. Apify sells Starter at $29 per month plus pay as you go, turns the subscription into spend credit, and lowers the compute unit rate on higher plans.

That last move is worth copying. When the unit rate drops as plans go up, the tiers reward commitment instead of just adding features. Customers who use more have a reason to move to a bigger plan rather than just paying more overage.

Usage products need a way to start without a purchase decision. A free allowance does that. 46.5% of the 989 catalog pages include a free tier, and for usage-based products it is usually a monthly quota rather than a feature-limited plan.

State the allowance in the same unit you bill. Algolia's Grow plan starts with 10K search requests per month free, then $0.50 per 1K. Fivetran states its free plan limits in billing units. Amplitude describes a generous free tier in volume terms, so buyers know exactly where the free plan ends.

Size the allowance so a real project can reach a working result on it, but a production workload cannot. That is the point where a customer has seen value and has a reason to pay.

A rate card answers 'how much per unit.' Buyers want to know 'how much per month.' Closing that gap is the single biggest improvement most usage pricing pages can make.

A calculator or slider does it best, and only 7.5% of catalog pages have one. GoCardless puts a fee calculator under its plan cards. Helicone runs a live pricing calculator. Bolna AI shows a live cost-per-minute calculator next to its plans. Hetzner Cloud shows hourly and monthly prices side by side so buyers do not have to multiply.

If a calculator is too much to build, write a worked example. Apify's 'How pricing works' section walks through a real bill. Blaxel shows the other side: a rate table priced to nine decimals and no worked example of a monthly bill. Every decimal place is accurate. None of them tell a buyer what they will spend.

Fear of a surprise bill is the main objection to usage pricing. The fix is to say exactly what happens at the limit.

The best pages put it in plain view. AgentQL prints overage rates in each plan card. GitHub Copilot defines each tier by a premium request quota and publishes the overage price. Chatfuel lists conversation volume per plan alongside the overage charge. Fathom Analytics spells out its overage rule on a page with a single price anchor of $15 per month.

Caps help too. DigitalOcean's hero promises monthly caps, which tells buyers the worst case before they read a single rate. If you can offer spend limits or alerts, say so on the pricing page, not just in the docs.

Then answer the follow-up questions in FAQs, which appear on 60.1% of catalog pages. For usage products the questions are predictable: what counts as a unit, what happens if I go over, can I set a cap, and do unused units roll over. Answer each one in a sentence or two, right below the rates.

Rate cards get long. Groq splits tables by modality. DigitalOcean prices per product, from App Platform at $0 to Kubernetes starting at $12 per month, with a sticky side nav for categories. Hetzner Cloud shows three server families with spec rows that make comparison easy.

Some rules that hold across the catalog:

• Lead with the starting number. Datadog and DigitalOcean use 'Starting at' labels so the first number a buyer sees is the entry point.
• Show the unit in terms people think in. Replicate shows every unit in both seconds and hours. Groq adds tokens per dollar in parentheses.
• Show real example prices. Replicate uses example model cards with real prices, which beats an abstract rate.
• Explain the unit in place. 7.7% of pages use more info tooltips on terms like compute units or premium requests.
• Keep a CTA near the prices. Replicate shows no plans, tiers or CTA near its prices, which leaves an interested buyer with nowhere to go.

Usage pricing is not always the answer. It works poorly when the unit does not track value, when customers need a fixed budget line, or when usage is so spiky that bills swing wildly month to month.

Collaboration tools are the clearest case. Value comes from people working together, not from a countable action, which is why Notion, Linear and Figma use per-seat pricing. Products sold to buyers with strict annual budgets also struggle, because procurement wants one number.

If your costs scale with usage but your buyers want predictability, use a middle path. Sell plans with included usage and a published overage, or sell prepaid credits. You keep your margin and the customer gets a bill they can plan around.

Usage pricing scales up revenue from big customers. It also gives them a reason to negotiate. Plan for that on the page.

Datadog notes multi-year and volume discounts for 500 or more hosts per month. Fivetran offers savings of up to 22% with an annual contract. HashiCorp repeats a $500 credit offer and lists three ways to buy at the top of the page. Each tells large buyers that a better rate exists without printing every discount tier.

For the very largest accounts, an enterprise tier with committed spend is the norm. It appears on 48.2% of catalog pages. Usage-based products tend to frame it as a commit and a lower rate rather than extra features.

If your usage model is sound but buyers still cannot work out their bill, the page is the problem. That is the kind of page Conversion Factory rebuilds in a pricing page revamp.