Directional Correctness Over Optimization¶

Why It Matters¶

Humans over-optimize prematurely, spending enormous effort perfecting solutions before knowing if they're solving the right problem. This feels productive but delays the learning that only comes from implementation. Directional correctness says: optimize only after you've validated you're in the right region of the solution space.

The Key Move¶

When tempted to optimize a solution, ask: "Do I have evidence this is even approximately right?" If not, choose the first good-enough option that lets you move forward and learn. Set a quality threshold of "good enough to test" rather than "as good as I can make it." Reserve optimization effort for later, after empirical validation that you're solving the right problem. Use the saved resources to test more alternatives instead of perfecting one.

Classic Application¶

Software development's "MVP" (Minimum Viable Product) approach. Rather than building full-featured version 1.0, ship the minimally functional version that creates user value. Twitter launched with just posting tweets - no replies, no retweets, no hashtags, no images. This was directionally correct (people want to share short updates) but nowhere near optimal. Shipping it revealed what to optimize (real-time updates, conversations) versus what wasn't needed (many initially planned features).

Surprising Application¶

Parenting strategies for new parents. New parents often agonize over finding the optimal approach to sleep training, feeding schedules, educational toys, etc. Directional correctness suggests: pick any reasonable approach (the range of "good enough" is wide), implement it consistently for two weeks, and observe results. Most approaches work adequately; the cost of choosing sub-optimally is low compared to the cost of decision paralysis. Optimize later with actual data about your specific child.

Failure Modes¶

Confusing "directional" with "careless" - shipping truly broken solutions that can't generate useful feedback. Using this as excuse for laziness rather than strategic under-optimization. Staying in directional mode forever, never optimizing when you have enough signal. Applying to domains where precision is required upfront (surgery, aerospace, legal contracts). Directional correctness in the wrong direction - moving confidently toward a bad solution.

Go Deeper¶

Reinertsen, The Principles of Product Development Flow, Chapter 3; Gawande, Better: A Surgeon's Notes on Performance (on "better versus best")