Re-measure 2 weeks after any constraint fix: if local metric improved but total throughput didn't, the constraint has shifted elsewhere
Two weeks after any constraint intervention, re-measure the old constraint and scan for queue buildup elsewhere; if the old constraint improved but total throughput did not improve proportionally, the constraint has shifted.
Why This Is a Rule
When you fix a constraint, the constraint doesn't disappear — it shifts to the next weakest link in the chain. Fixing the bottleneck in step 3 of your workflow may cause step 5 (which had excess capacity when step 3 was slow) to become the new bottleneck now that step 3 feeds it faster. If you only measure the old constraint ("Step 3 improved!") without measuring total throughput, you'll celebrate a local improvement that produced zero system improvement.
The two-week re-measurement serves two functions: verification (did the intervention actually fix the old constraint?) and detection (where did the constraint shift to?). Queue buildup is the diagnostic signal: if work is piling up at a new location in your workflow that previously had no queue, the constraint has shifted there. Two weeks provides enough data to distinguish a genuine constraint shift from temporary adjustment noise.
When This Fires
- Two weeks after any process improvement, tool upgrade, or workflow change
- When a local improvement didn't produce the expected overall results
- When Optimize only the single slowest step — improvements to non-bottleneck steps are wasted effort regardless of their magnitude's bottleneck-first optimization needs the follow-up verification
- Complements Optimize only the single slowest step — improvements to non-bottleneck steps are wasted effort regardless of their magnitude (optimize only the bottleneck) with the post-optimization verification cycle
Common Failure Mode
Celebrate-and-forget: "We fixed the bottleneck!" without re-measuring. The old bottleneck improved, but the new bottleneck now limits throughput at the same rate. Total output is unchanged despite the intervention.
The Protocol
(1) After any constraint intervention, set a calendar reminder for 2 weeks. (2) At 2 weeks, measure: Old constraint metric (did it actually improve?), Total throughput (did overall output improve proportionally?), Queue scan (is work piling up at any new location?). (3) If old constraint improved AND total throughput improved proportionally → success. The intervention worked and no new constraint has emerged. (4) If old constraint improved BUT total throughput didn't → the constraint has shifted. Identify the new constraint (look for queue buildup) and redirect optimization there. (5) If old constraint didn't improve → the intervention didn't work. Diagnose why and try a different approach before moving on.