If an approval step changed nothing in 20 instances, eliminate it as waste

Why This Is a Rule

Approval gates are installed to catch errors — but many outlive their usefulness while retaining their cost. An approval step that rubber-stamps twenty consecutive instances without requesting a single change is not catching errors. It's adding delay, creating bottlenecks, and consuming reviewer time without producing any value. The gate's historical rationale ("we added this after the 2019 incident") is irrelevant if the gate hasn't triggered in twenty consecutive passes.

Lean manufacturing calls this muda (waste): activity that consumes resources without adding value. In knowledge work, approval gates are particularly expensive because they create queues — work sits idle waiting for a reviewer who is busy with other tasks. The cost isn't just the reviewer's time; it's the delay cost multiplied by everything downstream that's waiting.

Twenty instances with zero changes is the evidence threshold. Below twenty, a pattern might not be established — the gate might catch errors rarely but importantly. At twenty consecutive passes with no changes, the gate is statistically demonstrated to be non-functional. Remove it.

When This Fires

• During process audits or workflow optimization
• When a review step consistently adds delay without adding feedback
• After noticing that approval queues are a recurring bottleneck
• Any time someone says "we've always had this review step" without evidence it's catching anything

Common Failure Mode

Keeping the gate because of its origin story: "We added this after the incident in 2019." The incident was real, the gate was a reasonable response — but if the gate hasn't caught anything since, either the underlying cause was fixed by other means or the gate's design doesn't actually catch the problem. The origin story justifies the gate's existence emotionally, not functionally.

The Protocol

For each approval step in your process: (1) Track the last 20 instances. How many resulted in a change request from the reviewer? (2) If zero → eliminate the gate. Move the review to a sampling basis (spot-check 1 in 10) if you want a safety net. (3) If 1-3 → the gate has marginal value. Consider whether the changes it caught could be caught earlier in the process. (4) If 4+ → the gate is functional and earning its cost. Keep it.