Optimize only the single slowest step — improvements to non-bottleneck steps are wasted effort regardless of their magnitude

Why This Is a Rule

Goldratt's Theory of Constraints provides perhaps the most counter-intuitive insight in operations management: improving a non-bottleneck step cannot improve overall throughput. If your workflow has five steps taking 10, 20, 30, 15, and 10 minutes respectively, the throughput is governed entirely by the 30-minute step. Making the 20-minute step twice as fast (saving 10 minutes) produces zero improvement in total cycle time — work still queues at the 30-minute bottleneck.

This is counter-intuitive because local improvements feel productive. "I saved 10 minutes on step 2!" But the 10 minutes were saved in a non-binding constraint. The work still waits 30 minutes at the bottleneck. The only improvements that propagate to system-level output are improvements to the bottleneck itself. Everything else is local optimization that produces zero global benefit — the operational equivalent of rearranging deck chairs.

Applied to personal workflows, this means resisting the urge to optimize whatever is convenient, enjoyable to improve, or obviously imperfect. The only valid optimization target is the measured bottleneck (Measure actual elapsed time over 3-4 cycles before optimizing — felt difficulty systematically misidentifies bottlenecks). This requires discipline because non-bottleneck improvements are easier to see and more satisfying to implement — they produce visible local gains that mask their global irrelevance.

When This Fires

• After measuring workflow step times (Measure actual elapsed time over 3-4 cycles before optimizing — felt difficulty systematically misidentifies bottlenecks) and identifying the bottleneck
• When you're tempted to optimize a non-bottleneck step because it seems "easy to improve"
• When workflow optimization efforts produce local improvements but no overall time reduction
• When prioritizing which process improvement to tackle first

Common Failure Mode

Spray-and-pray optimization: improving multiple steps simultaneously under the assumption that all improvements add up. They don't. Only bottleneck improvements compound into throughput gains. Non-bottleneck improvements create excess capacity that sits idle upstream of the bottleneck or creates congestion downstream of it.

The Protocol

(1) Measure all workflow steps (Measure actual elapsed time over 3-4 cycles before optimizing — felt difficulty systematically misidentifies bottlenecks) and identify the single slowest step. (2) Direct 100% of optimization effort at this step. Ignore all other steps regardless of how obviously improvable they are. (3) After improving the bottleneck, re-measure the entire workflow. The bottleneck may have shifted to a different step. (4) Repeat: identify new bottleneck → optimize → re-measure. This cycle continues until the workflow is fast enough or further optimization has diminishing returns. (5) If you catch yourself optimizing a non-bottleneck step, stop immediately. Ask: "Is this the measured bottleneck?" If no, it's wasted effort. Redirect to the actual constraint.