Learn it → Apply it → Save as Skill
A number you can see starts to feel more real than a truth you can't. The McNamara Fallacy is what happens when a team trusts an easy metric so completely that everything the metric fails to capture quietly stops counting, and eventually stops being discussed at all.
Dr. Wallace's Favorite Mental Models turns useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
Easy numbers crowd out hard truths. A team measures what's convenient, treats the inconvenient as unimportant, and eventually forgets the inconvenient was ever part of the goal.
Important: step one is ordinary good practice. The fallacy is steps two through four, where neglect quietly becomes denial.
The model is most useful whenever a dashboard, score, or KPI has quietly started standing in for the thing it was only ever meant to track.
Each step feels reasonable in isolation. The damage comes from how naturally one step slides into the next.
Track whatever can be counted cleanly: calls handled, test scores, body counts, click-through rates.
Whatever resists easy measurement gets a weak proxy, a low-response survey, or no number at all.
Because it has no clean number, leadership starts treating the hard-to-measure part as a minor concern.
The hard-to-measure part disappears from the report entirely, as if it were never part of the goal.
Open a use case to see what the model changes.
A team optimizes average handle time, response count, or units shipped, while the underlying goal, a satisfied customer or a working product, is no longer separately checked.
Standardized scores are easy to compare across schools. Curiosity, judgment, and the ability to write a real argument are harder to score, so they quietly get less attention than the test.
A model posts a strong accuracy or benchmark number while failing on messy, underrepresented, or adversarial inputs that the benchmark never sampled. The score looks finished; the product is not.
A hospital is scored on 30-day readmissions. Staff can hit the number by discharging earlier, while patient confidence, pain, and recovery quality go uncounted.
As U.S. Secretary of Defense, McNamara's staff reported enemy body counts as a precise measure of progress in Vietnam. Morale, local political sentiment, and the reliability of the counts themselves were much harder to quantify, and were steadily treated as secondary.
A job offer is compared almost entirely on pay, because pay is a clean number. Commute stress, autonomy, and whether the work is meaningful are real factors that resist a tidy comparison.
This is an illustrative example designed to teach the method.
Expand for the distinction between the coined term, the man it was named after, and the four-step structure.
Social scientist Daniel Yankelovich described the pattern in a 1971 analysis of quantification in American life and policy. He later named it after Robert McNamara, whose management style he saw as the clearest example of it.
As U.S. Secretary of Defense from 1961 to 1968, McNamara brought a statistics-heavy management style from the auto industry into the Pentagon. During the Vietnam War, enemy body counts became a central measure of progress, even though they were difficult to verify and said little about political will or long-term outcomes. Yankelovich argued this contributed to a strategic misjudgment of the war.
As Yankelovich stated them: measure whatever can easily be measured; disregard what can't easily be measured, or give it an arbitrary quantitative value; presume that what can't be measured easily isn't very important; and conclude that what can't be easily measured doesn't really exist. Each step is a small, reasonable-sounding move. The sequence is what causes the damage.
No number on the dashboard is read as "no problem," when it may only mean "no one built a way to measure it yet."
A measure gets adopted because it's cheap to collect, not because it is the best available proxy for the goal.
Average handle time, test scores, and body counts are proxies. The real goals, trust, learning, and a durable peace, are not the same thing as the number standing in for them.
A 4% response rate survey is still a number, but a small, biased sample is not automatically more reliable than careful qualitative judgment.
When raising "the number looks good but something feels off" gets treated as a complaint rather than a signal, the gap gets reported less, not solved.
A measure that fit the goal two years ago may have drifted. Teams rarely schedule a moment to ask whether it still does.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply the McNamara Fallacy, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
It's the shadow side of that saying. "What gets measured gets managed" is a reason to measure. The McNamara Fallacy is the warning that comes right after: what doesn't get measured can quietly stop being managed, or even stop being seen as part of the job.
No. The fix is not fewer numbers. It is pairing the clean numbers you have with a deliberate, recurring check on what they leave out, through qualitative review, direct observation, or a named owner for the parts that resist measurement.
Goodhart's Law describes a metric being actively gamed once people optimize for it. The McNamara Fallacy describes something quieter: the parts of the goal that were never captured by any metric simply fading out of the conversation. A team can fall into this fallacy without anyone gaming anything.