PlanWright
First-Pass Acceptance Rate
Home/Docs/Signals & Insights/First-Pass Acceptance Rate

Signals & Insights

First-pass acceptance rate

Insights · Workspace

The flagship metric for objective quality — and for the quality of human direction.

Overall rate

The percentage of completed objectives accepted without any rejections. The single most diagnostic number in the system.

Healthy

Rate is high and stable, or rising. Agents execute cleanly against well-written objectives. Human direction quality compounds: each well-scoped objective makes the next easier to write.

Watch for

Rate is low or declining. Almost always an objective quality issue, not an agent issue. Ask: were acceptance criteria explicit? Did the objective describe an outcome or an implementation?

Trend vs. prior period

First-pass rate compared against the equivalent prior window (7-day vs. prior 7 days, 30-day vs. prior 30 days).

Healthy

Rate is stable or improving across consecutive windows. The team is getting more deliberate about defining objectives before scheduling them.

Watch for

Rate declining across multiple windows — the team is scheduling objectives that aren't ready. Usually a sign that urgency is overriding definition quality.

Per-agent breakdown

First-pass rate and average rejection cycles broken down by agent kind.

Healthy

Rates are consistent across agents for the same class of work, or the team has learned which agents work best for which objective types and routes accordingly.

Watch for

One agent kind consistently underperforming on a class of work. Usually means those objectives need more context for that agent — file pointers, interface contracts, explicit criteria — not a different agent.

Clarity-flag correlation

Whether objectives that agents flagged as unclear before execution have lower first-pass rates.

Healthy

Clarity flags are rare. When raised, the team refines the objective before running. First-pass rates on flagged objectives, once resolved, are comparable to unflagged ones.

Watch for

High flag rate, or flags consistently overridden and objectives run anyway. An overridden clarity flag is the strongest predictor of rejection — the agent couldn't interpret the objective, so the first attempt is almost certain to miss.