EAPM — Effective Actions Per Minute

EAPM is APM with errors subtracted. It counts only successful, productive actions — giving you a clearer picture of how much useful work is being accomplished per minute.

Definition

Effective Actions Per Minute strips out failed tool executions from the action count, leaving only the actions that actually produced results.

Formula:

EAPM = (tool_uses + prompts - errors) / session_duration_minutes

Data source

EAPM uses the same deterministic data as APM, plus error counts:

• tool_uses: Counted from PostToolUse hook events
• prompts: Counted from UserPromptSubmit hook events
• errors: Counted from PostToolUseFailure hook events — fired when a tool execution fails
• Duration: From started_at to ended_at timestamps on plugin_session_summaries

Like APM, there is no estimation or sampling. Every value is derived from actual recorded events.

The APM-EAPM gap

The difference between APM and EAPM reveals your error rate. This gap is one of the most actionable signals in the Control Center.

APM	EAPM	Error Rate	Interpretation
10	9.5	5%	Healthy — minimal errors
10	7	30%	Notable — investigate recurring failures
15	8	47%	High — nearly half of all actions are failing
20	5	75%	Critical — most effort is wasted on retries

If APM is 15 but EAPM is 8, that's a 47% error rate. Common causes include:

• Permission issues: File or directory access errors causing repeated tool failures
• Missing dependencies: Build or test commands failing because of missing packages
• Incorrect paths: File read/write operations targeting non-existent paths
• Prompt ambiguity: Vague instructions leading Claude to attempt actions that fail, then retry with corrections

Why it matters

EAPM separates productive work from noise. A session with high APM and low EAPM is spending most of its time failing and retrying — the high action count is masking low productivity.

Consider two sessions:

• Session A: APM 8, EAPM 7.5 — Steady, efficient work with a 6% error rate
• Session B: APM 20, EAPM 6 — Frantic activity with a 70% error rate

Session A is almost certainly more productive despite the lower APM. Session B is burning through tokens and time on failed attempts.

When you see a persistent gap between APM and EAPM, investigate:

1. Check which tools are failing (the Usage tab in the Control Center shows tool-level breakdowns)
2. Look for patterns — is it always Bash commands? File operations? Specific directories?
3. Address the root cause: fix permissions, install dependencies, or refine your prompts

The StarCraft analogy

In competitive StarCraft, "Effective APM" removes spam clicks — rapidly clicking the same command or clicking on empty ground. Only actions that actually change the game state count. A player with 300 APM but 100 EAPM is mostly clicking without purpose. The same principle applies here: only productive tool executions count toward your effective rate.

Example

A 15-minute session with 5 prompts, 60 tool executions, and 12 errors:

APM  = (60 + 5) / 15 = 4.3
EAPM = (60 + 5 - 12) / 15 = 3.5
Error rate = 12 / 65 = 18%

The 0.8-point gap suggests moderate error rates — worth monitoring but not alarming.