AboutHow we built thisSponsorshipShop
SearchSubscribeDecision ToolsBusiness ModelsFrameworksReading Lists
Privacy PolicyTerms of UseCookie PolicyRefund PolicyAccessibilityDisclaimer

© 2026 Faster Than Normal. All rights reserved.

Faster Than Normal
PeopleBusinessesShopNewsletter
Start reading →
  1. Home
  2. Guides
  3. The Map is Not the Territory: Why All Models Are Wrong
Guide

The Map is Not the Territory: Why All Models Are Wrong

Alfred Korzybski's insight that models and reality are fundamentally different things — and confusing the two is the source of most catastrophic errors in business and life.

In this guide

  1. A menu is not a meal
  2. All models are wrong, but some are useful
  3. When the map caused a financial crisis
  4. Metrics become targets and then they lie
  5. Why experts are especially vulnerable
  6. Keeping one foot in the territory

A menu is not a meal

In 1931, Polish-American scholar Alfred Korzybski presented a paper to the American Association for the Advancement of Science in which he introduced a phrase that would outlive the rest of his work: 'the map is not the territory.' His argument was that human beings navigate reality through abstractions — language, models, categories, frameworks — and that these abstractions are not reality itself. They are representations. And every representation, no matter how detailed, leaves something out. The examples are everywhere once you see the pattern. A restaurant menu describes dishes but cannot feed you. A financial model projects returns but cannot guarantee them. A resume describes a candidate but cannot predict how they will perform under pressure. An org chart shows reporting lines but says nothing about who actually has influence. A photograph of a sunset is not a sunset. Korzybski's insight was not merely philosophical. He argued that confusing the map with the territory — mistaking your representation of reality for reality itself — was the root of most human error. When you act on a model as though it were the world, you inevitably encounter gaps the model didn't account for. The question is whether those gaps are trivial or catastrophic. In business, in medicine, in government, the answer has too often been catastrophic.

All models are wrong, but some are useful

The British statistician George Box captured the map-territory distinction in a single sentence: 'All models are wrong, but some are useful.' A map of London that included every brick, pipe, wire, and blade of grass would be useless — it would be as large and complex as London itself. The entire value of a map comes from what it omits. Simplification is the point. But useful simplification creates a specific danger: the more useful a model is, the more tempting it becomes to treat it as truth. A weather forecast that is right 90% of the time trains you to trust it — and the one time it is catastrophically wrong, you are caught without an umbrella. Or, less trivially, without an evacuation plan. Box's warning applies to every model in every domain. Economic models are wrong — they simplify human behaviour into mathematical functions. Strategy frameworks are wrong — they compress messy competitive dynamics into two-by-two matrices. Business plans are wrong — they project a single future when the actual future is a probability distribution. The wrongness is not a flaw. It is a design feature. But the moment you forget that your model is a deliberate simplification, you begin to make decisions as though the simplification were a fact. That is when the trouble starts.

When the map caused a financial crisis

The 2008 financial crisis is perhaps the most expensive map-territory confusion in history. The maps in question were the quantitative risk models used by banks, rating agencies, and regulators to assess mortgage-backed securities. These models made assumptions: that housing prices would not decline nationally, that individual mortgage defaults were largely uncorrelated, and that markets would remain liquid enough to allow orderly selling during downturns. Each assumption was a simplification — a feature of the map, not of the territory. Housing prices had, in fact, declined regionally many times. Mortgage defaults were correlated because they were driven by the same economic conditions. And market liquidity, as any crisis demonstrates, evaporates precisely when you need it most. But the models worked beautifully during the period they were tested against. They fit the historical data. They produced elegant risk assessments. Rating agencies used them to stamp AAA ratings on securities that were, in retrospect, toxic. Banks used them to justify leverage ratios that left zero margin for error. Regulators used them to reassure themselves that the system was sound. When the territory diverged from the map — when housing prices fell, defaults correlated, and liquidity vanished simultaneously — the result was a global financial meltdown. The models had not lied exactly. They had simplified, and the people using them had forgotten what the simplifications left out.

Metrics become targets and then they lie

British economist Charles Goodhart observed in 1975 that 'when a measure becomes a target, it ceases to be a good measure.' This is map-territory confusion in organisational form, and it is ubiquitous. A metric starts as a useful proxy — a map — for something you care about in the real world. Revenue is a map of value creation. Test scores are a map of learning. Patient satisfaction scores are a map of healthcare quality. As long as the metric remains a map, it is useful. The trouble begins when the metric becomes the objective. A hospital that optimises for patient satisfaction scores may start avoiding difficult conversations that patients need to hear. A school that optimises for standardised test scores may narrow its curriculum until students can pass the test but cannot think. A company that optimises for quarterly revenue may sacrifice customer relationships, product quality, and employee trust — all of which are invisible to the metric but essential to the territory. Goodhart's Law is a specific instance of the broader map-territory problem: the metric was a useful representation of something real, but once the representation became the target, people optimised the map at the expense of the territory. They hit the numbers and missed the point. In every organisation, the question to ask is: are we managing the metric, or managing the thing the metric is supposed to measure?

Why experts are especially vulnerable

Deep expertise creates increasingly sophisticated maps — and increasing temptation to mistake them for territory. The economist who builds elegant models may forget that economies contain human beings who panic, follow fads, and act against their own interests. The military strategist who creates beautiful war plans may forget Helmuth von Moltke's warning that no plan survives first contact with the enemy. The data scientist who builds accurate predictive models may forget that accuracy measured on historical data says nothing about accuracy in a future that behaves differently. Nassim Nicholas Taleb calls this the 'ludic fallacy' — treating real-world uncertainty as though it follows the clean, well-defined rules of a casino game. In a casino, the probabilities are known, the payoffs are defined, and the rules do not change. In reality, the probabilities are unknown, the payoffs are uncertain, and the rules shift without warning. The expert's map is a casino model. The territory is something far wilder. The most dangerous maps are the ones that are mostly right. A model that is obviously wrong gets discarded. A model that is right 95% of the time earns trust — and that trust means the 5% of cases where it fails arrive as complete surprises. The expert's confidence in the map is precisely what prevents them from noticing when the territory has shifted.

Keeping one foot in the territory

The antidote to map-territory confusion is not abandoning maps — they are essential for navigating complexity. The antidote is building habits and systems that force you to check your maps against reality, regularly and ruthlessly. Jeff Bezos describes Amazon's approach as 'high-velocity decision-making': make decisions based on your best model, but build fast feedback loops so you discover map-territory gaps quickly, before they compound. The mechanism matters more than the intention. It is not enough to believe your maps might be wrong. You need structural practices that reveal when they are. Visit customers instead of just reading NPS surveys. Walk the factory floor instead of just reviewing dashboards. Talk to front-line employees instead of just reading management reports. Every layer of abstraction between you and reality is a place where the map can silently diverge from the territory. The CEO who relies exclusively on executive summaries is navigating by a map of a map of a map. Charlie Munger recommends systematically looking for disconfirming evidence — information that contradicts your current model. This is psychologically difficult because humans prefer confirmation, but it is the single most effective way to find map-territory gaps before they become crises. The discipline is simple to describe and hard to practice: treat every model, including your most trusted ones, as provisionally useful rather than permanently true.

Related mental models

first principles thinkinggoodharts lawcircle of competenceskin in the gamefeedback loops

Continue exploring

FT

Mental model

First Principles Thinking

GL

Mental model

Goodharts Law

CC

Mental model

Circle Of Competence

SG

Mental model

Skin In The Game