Margin of Error

A margin of error is the slack CF folds around a breakpoint and around a measurement, because neither is perfectly precise. Asking how much food is enough not to be hungry, or whether furniture fits through a door, you don’t know the exact threshold—but you know it approximately, and every measurement carries its own error band too. CF tracks both: a margin around the breakpoint and a margin around the data point.

The payoff is robustness. CF’s goals aim for enough of a factor plus a margin, not for a maximum, so a pass/fail judgment doesn’t flip on minor fluctuations. Temple argues that margins are typically much smaller than the categories they sit inside, so most data points land far outside every breakpoint’s margin and can be classified easily and accurately. We remember the close calls only because of selective focus—we attend to borderline cases and ignore the many easy ones.

This grounds CF’s case against weighted-factor scoring. Because a working system carries margins (tolerances, resilience) on most factors, slight quantitative changes fall inside the margin and shouldn’t move your verdict at all. Rewarding such changes with a higher score is an error: extra excess capacity beyond the margin doesn’t merit more credit. When a margin genuinely does matter—a solution barely works—CF’s advice is to seek a non-close-call alternative, shrink the margin by measuring or understanding the breakpoint better, or abandon a project you can only barely do. The link to Theory of Constraints is direct: design systems with deliberate margins so minor variation never demands attention.