OpsVis.ai
Use CasesProductPricingIntegrationsGovernmentContact
OpsVis.ai
Sign InSign Up

Margin leakage

Margin leakage is the gradual, often unnoticed erosion of gross margin caused by a gap between what a company is actually billed by its vendors and what it charges its customers — for example, unbilled overage, a stale rate card, or a vendor price change that never gets reflected downstream.

What causes margin leakage in AI billing

Margin leakage in AI-SaaS billing has a few recurring shapes: a model vendor raises its per-token price and the customer-facing rate card is not updated to match; a customer's usage crosses into a higher, more expensive vendor pricing tier that the company's own rate card never accounted for; usage is metered incorrectly (undercounted internally, so a customer is undercharged relative to real cost); or a negotiated vendor discount expires or fails to apply, quietly raising the true cost basis behind an unchanged customer price.

Each of these is small on any single invoice and easy to miss in isolation. The reason margin leakage is dangerous is that it compounds silently across every customer and every billing cycle until it shows up as an unexplained gap between reported and actual gross margin.

Catching it before it compounds

Detecting margin leakage requires comparing actual vendor cost (not list price) against actual customer billing at the same granularity — per customer, per metric, per period — which is what vendor cost reconciliation and cost attribution are built to do. A margin monitoring process that only looks at aggregate revenue and aggregate vendor spend can miss leakage that is fully offset in the mix but is actively destroying margin on specific customers or product lines.

Related terms

  • AI gross margin
  • Vendor cost reconciliation
  • Blended cost vs. list price
  • Cost attribution

Frequently asked questions

What causes margin leakage in AI billing specifically?

The most common causes are vendor price increases that never propagate to the customer-facing rate card, usage that crosses into a more expensive vendor pricing tier the company didn't account for, undercounted internal usage metering, and negotiated vendor discounts that lapse without anyone noticing.

How is margin leakage different from margin compression?

Margin compression is an expected, structural narrowing of margin — for example, because AI inference is a variable cost that does not shrink with scale the way fixed infrastructure cost does. Margin leakage is unintended and correctable: it comes from a billing, pricing, or reconciliation gap rather than the underlying cost structure of the product.

Built by OpsVis.ai. The source code is available on GitHub.