The useful lie of every model
Statistician George Box wrote: 'All models are wrong, but some are useful.' A map of London that included every brick, pipe, and blade of grass would be useless — it would be as complex as London itself. The value of a map is precisely what it leaves out. But this useful simplification becomes dangerous when you forget what was omitted. The 2008 financial crisis was, at its root, a map-territory confusion: risk models treated mortgage-backed securities as though the model's assumptions (uncorrelated defaults, liquid markets) were features of reality rather than simplifications.
Metrics become targets and then they lie
Charles Goodhart observed that 'when a measure becomes a target, it ceases to be a good measure.' This is map-territory confusion in organisational form. A hospital that optimises for patient satisfaction scores may start avoiding difficult truths that patients need to hear. A company that optimises for quarterly revenue may sacrifice long-term customer relationships. The metric was a map of something real — patient health, business health — but once the map became the objective, it diverged from the territory it was meant to represent.
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 act irrationally. The strategist who creates beautiful frameworks may forget that competitors don't read their memos. Nassim Taleb calls this the 'ludic fallacy' — treating real-world uncertainty as if it follows the clean rules of a casino game. The most dangerous maps are the ones that are mostly right, because the user stops checking them against reality.
Keeping one foot in the territory
The antidote isn't to abandon maps — they're essential for navigating complexity. The antidote is to continuously verify your maps against reality. Jeff Bezos calls this 'high-velocity decision-making': make decisions based on your best model, but build in fast feedback loops so you discover map-territory gaps quickly. 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 reports. Every layer of abstraction between you and reality is a place where the map can silently diverge from the territory.