Apply the 37% rule to search decisions — explore and reject the first third, then commit to the first option that beats the best explored

Why This Is a Rule

The optimal stopping problem (also called the secretary problem) has a mathematically proven solution: when selecting from a pool of candidates evaluated sequentially, the strategy that maximizes the probability of selecting the best candidate is to reject the first 37% (1/e) unconditionally, then accept the next candidate who exceeds all previously seen candidates. This gives approximately a 37% chance of selecting the absolute best option — which, counterintuitively, is the highest achievable probability for this class of problems.

The exploration phase (first 37%) serves a pure calibration function: you're learning what "good" looks like without committing. Without this phase, you risk committing too early to a candidate who seems good only because you lack a comparison baseline. With it, you've seen enough of the distribution to know when a genuinely excellent option appears.

This applies to hiring, apartment hunting, dating, vendor selection, and any sequential-evaluation decision where you can't go back to previously rejected options. The 37% is a mathematical constant, not a heuristic — it's provably optimal under the problem's assumptions.

When This Fires

• When making sequential-evaluation decisions from a known or estimable pool size
• When hiring from a candidate pipeline with a known number of applicants
• When apartment hunting with a finite number of viewings scheduled
• When the decision structure is: evaluate options one at a time, accept or reject, can't return to rejected options

Common Failure Mode

Committing during the exploration phase because an early candidate seems excellent: "This person is amazing — what if I don't find anyone better?" The 37% rule says: reject them anyway. You're in the calibration phase. The "amazing" assessment lacks the baseline that only the full exploration phase provides. What feels like the best candidate at 15% of the pool may be merely the first competent one you've seen.

The Protocol

(1) Estimate the pool size: how many candidates/options will you evaluate? (2) Calculate 37% of that number (round to nearest integer). This is your exploration phase. (3) During exploration: evaluate each option carefully, note quality, but reject all of them. No exceptions. (4) After exploration: commit to the first option that exceeds the best you saw during exploration. (5) If no option exceeds the exploration-phase best → you're forced to take the last option or restart. This happens ~37% of the time and is the known cost of the strategy. (6) Adjust for real-world constraints: if options can be recalled, if pool size is uncertain, or if evaluation is imperfect, the optimal fraction shifts lower (20-30%). The 37% assumes the pure mathematical case.