Attribution & Finance
Metering accuracy
How completely and correctly a system counts the usage it bills. A meter is accurate when every unit of consumption is captured exactly once — idempotently, under load, attributed to the right customer. The maxim: metering accuracy is revenue accuracy — every dropped or double-counted event is money mis-billed.
Example
A batched meter drops events during a traffic spike; each dropped event is a token served but never billed. An idempotent meter keyed on your own request id captures each call exactly once, so retries never double-bill and spikes never under-bill.
This is a Tokenality concept. See how it works in the product overview or the live playground.
Related terms
Revenue leakage
Revenue a business could bill for but doesn't — because of a pricing proxy, an un-billed overage, a dropped metering event, or a missed reconciliation. For AI resellers it's the gap between usage-based cost and proxy-based price; industry estimates put it at 3–9% of revenue for usage-based software.
Usage-based billing
Charging customers in proportion to what they actually consume — tokens, calls, or compute — instead of a flat fee, seat count, or user-count proxy. It aligns price with cost: because AI cost scales with tokens, usage-based billing is the only pricing that keeps a reseller's margin stable as customers grow.