When the cost of late error discovery exceeds 5x early discovery, invest more in detection before proceeding to the next phase

Why This Is a Rule

Software engineering has long understood that bugs found in design cost 1x to fix, bugs found in development cost 10x, and bugs found in production cost 100x. This cost multiplication applies to any multi-phase commitment: errors discovered early are cheap to fix because little has been built on top of them. Errors discovered late require rework of everything built on the flawed foundation.

The 5:1 ratio is the decision threshold: when discovering an error now would cost X, but discovering it after the next phase would cost 5X or more, the expected value of additional detection investment is strongly positive. Even if the detection effort only has a 20% chance of finding an error, the investment is worthwhile because 0.2 × 5X savings > the detection cost.

This transforms the vague advice "test early" into a quantitative decision rule. At each phase boundary, you calculate the cost multiplier and let it determine how much detection effort to invest. Low ratios (2:1) justify proceeding quickly. High ratios (10:1, 50:1) justify pausing for thorough verification — even if it feels slow.

When This Fires

• At every phase boundary of a multi-phase project (design → build → test → deploy)
• Before committing to the next phase of any staged investment
• When deciding how much testing, review, or validation to invest at each checkpoint
• Complements Test your riskiest assumptions first and cheapest — list 3-5 that would invalidate the most work if wrong (assumption testing) with ongoing phase-gate detection

Common Failure Mode

Rushing through phase gates because "we're behind schedule": "Let's skip the design review and go straight to implementation." If the cost ratio is 10:1 (design errors caught now cost $10K, caught in production cost $100K), skipping the review saves days but risks $90K in unnecessary rework. The schedule pressure is real but the cost asymmetry is larger.

The Protocol

(1) At each phase boundary, estimate two costs: Cost of finding an error now: what does it cost to discover and fix a foundational error at this stage? Cost of finding the same error after the next phase: what would it cost if the error survived undetected through the next phase? (2) Calculate the ratio: late cost ÷ early cost. (3) If ratio < 5:1 → proceed normally. Detection investment at this stage has low leverage. (4) If ratio ≥ 5:1 → increase detection before proceeding: additional reviews, prototypes, tests, expert consultations, or validation experiments. The detection investment is justified by the cost asymmetry. (5) The higher the ratio, the more detection to invest. A 50:1 ratio justifies pausing for weeks if necessary. A 5:1 ratio justifies a few extra days of verification.