Negative, extreme, or boundary case

Go inspect the failure, the outlier, the dropout, or the edge condition directly, because that is where the model breaks.

Your eating plan works Monday through Thursday, then collapses most Fridays. You average the week, see "mostly fine," and conclude you need more willpower. The averaging is the mistake. The three collapses carry more information than the 25 compliant days, because the collapses are where your model of yourself is wrong.

This move says go inspect the failure, the outlier, the dropout, or the edge condition directly, because that is exactly where the plan breaks. The cue is noticing you keep summarizing your record and skimming past the exceptions. Steps: pick the worst specific instances (the 11pm Friday order, the workouts that vanished during travel weeks), reconstruct each one in concrete detail, and compare conditions: what was different about sleep, stress, location, timing. Also check quiet dropouts, the habits that disappeared without a decision. When it works, "I lack willpower" becomes "I skip lunch on Fridays and then overshoot at night," which is fixable by Friday. Don't use it when the extreme case is a genuine one-off; rebuilding your whole plan around a single freak event overfits, and the typical days still set the baseline.