Skip to content

Article / Evaluation reporting

Why failed AI agent runs belong in evaluation denominators

Removing timeouts, invalid tool calls, abandoned tasks, or terminal failures can make an agent look more reliable by excluding the very behavior the evaluation was meant to measure.

For
Teams reporting completion, reliability, and tool-use rates across agent runs
Outcome
Define denominators that preserve failures and disclose legitimate exclusions

Answer first

Start from every eligible run, then classify what happened

If a run entered the declared protocol, keep it in the primary denominator unless a predeclared eligibility rule proves that the environment—not the agent—made the run unusable.

A timeout, malformed call, repeated conflict, or voluntary stop can be the outcome of interest. Deleting it after the fact changes the question from ‘How often did this agent complete?’ to ‘How often did it complete among runs that completed enough for us to score?’

That narrower statistic may still be useful, but it needs its own name and denominator. It should not replace the all-eligible-run result.

Run accounting

Separate agent failures, environment failures, and missing evidence

Run classifications and default reporting treatment
ClassificationExamplePrimary denominator
CompletedReached a terminal authoritative stateInclude
Agent failureInvalid calls, abandonment, retry exhaustionInclude
Protocol stopDeclared action or time budget reachedInclude and label
Environment failureService defect prevents a valid runReport separately; apply predeclared rule
Missing evidenceRecord cannot establish what happenedDisclose; do not silently drop

Metrics

Different measures need explicit, defensible denominators

Completion rate can use eligible runs, legal-action rate can use attempted actions, and policy adherence can use decisions with sufficient decision-time evidence.

Those populations are not interchangeable. A run with no accepted actions still belongs in a completion denominator, even though it contributes no accepted transition to an outcome distribution. A rejected attempt belongs in the attempted-action denominator even when the agent later repairs it.

Publish the numerator and denominator beside every rate. Counts make exclusions and small samples visible in a way that a rounded percentage cannot.

Integrity

Declare exclusion rules before seeing which variant benefits

A fair comparison applies the same failure and exclusion policy to every model, prompt, or strategy under test.

  • Name the eligible run population before execution.
  • Record every started run and its terminal classification.
  • Publish environment failures rather than hiding them inside agent results.
  • Show sensitivity when a reasonable alternative exclusion rule changes the conclusion.

Next step

Read the evaluation methodology

Design a study that keeps objective evidence separate from reporting choices.

Read the evaluation methodology