Skip to content
The Product Guys
All patterns
Money6 min read

Usage Limits and Metering

Pricing tied to a counted unit, with the count shown before it runs out.

The problem it solves

Some products cost more to run the more they are used. Flat pricing either overcharges light users or loses money on heavy ones, and a limit nobody can see feels like a punishment.

Two ways to hit the same limit

100%75%50%25%0%Warning thresholdWarned at 75%Unwarned123456Week of the billing periodShare of quota used
Both accounts reach the cap in the same week. The one warned at seventy-five percent has time to act; the one told only at the stop loses work it had already scheduled.

Metering aligns what a customer pays with what they consume. It also transfers uncertainty from you to them, and people dislike uncertainty in a bill more than they dislike a higher price. Everything in the design of a meter is about giving that uncertainty back.

Choosing the unit

  • The unit should track the value received, not the cost incurred. Charging per API call when the customer cares about resolved tickets invites them to optimise against you.
  • It should be countable by the customer without your dashboard. If they cannot estimate next month's bill on the back of an envelope, they will budget for the worst case and buy less.
  • It should not punish good behaviour. A meter on saved items discourages saving, which is usually the habit you wanted.
  • It should be stable. Redefining the unit after adoption is a price rise dressed as a clarification, and customers read it that way.

Then decide what happens at the line. Hard stops protect your costs and break the customer's workflow. Soft overage keeps the work running and produces a bill nobody approved. The usable middle is a soft stop with a clear resume: pause the work, tell them exactly what happened, and give a one-click way to continue.

Limit copy that panics people

  • Quota exceeded.
  • You have reached your limit. Upgrade to continue.
  • Error 429: rate limit

Limit copy that keeps them working

  • You have used 9,400 of 10,000 credits this month. Your credits reset on 1 October.
  • This job is paused because it needs 800 credits and you have 600 left. Add credits or wait until 1 October.
  • Too many requests. Retrying in 30 seconds. Your plan allows 100 requests per minute.

Show the meter continuously, not at the moment of failure. A number in the account area, a nudge: A change to how choices are presented that shifts behaviour without removing any option. at seventy-five percent, and an email at ninety percent cost almost nothing and prevent the two worst outcomes: a surprise bill and a job that died overnight.

Where metering turns predatory

Metering crosses into dishonesty when the count is unauditable, when overage rates are buried outside the pricing page, or when unused credits expire silently at the period boundary while the seller keeps the money. Automatic top-ups that recharge a card without a per-charge notice are close to forced continuity, and consumer protection rules in several jurisdictions now expect explicit consent for each recurring charge and an equally easy way out.

When it fits

  • Serving cost genuinely scales with a countable unit, such as compute, storage, or messages sent.
  • Customer value scales with the same unit, so a bigger bill arrives alongside a bigger result.
  • You can show the count in the product in real time or close to it.
  • Usage is predictable enough month to month that a finance team can budget for it.

When it backfires

  • The unit is opaque, such as an internal credit whose exchange rate changes without notice.
  • Usage is spiky, so a single bad week produces a bill that ends the relationship.
  • The meter discourages exactly the behaviour that creates retention, like inviting teammates or storing history.
  • Overage is silent and automatic, which turns a pricing model into a trap and increasingly into a regulatory problem.

Products using it

  • Amazon Web Services

    Meters dozens of separate units and offers budgets and alerts, which exist because the model is otherwise impossible to forecast.

  • Twilio

    Prices per message and per minute, with rates published openly so a customer can calculate a bill before signing up.

  • Vercel

    Shows usage against plan limits in the dashboard and warns as a project approaches them rather than only on failure.

The psychology under it

The takeaway

Pick a unit the customer can count themselves, then show the count long before it runs out.

Finished the teardown? Bank it and the day counts toward your run.