How to estimate anything

You will never know how many piano tuners work in Chicago. You can still get within a factor of two — and the technique that gets you there is learnable in about ten minutes.

Zoggle · a guide for people who like numbers

Here is a question you cannot answer: how many gallons of paint would it take to paint every house in America? You don't know. Nobody walking down the street knows. There is no reasonable expectation that this fact lives in your head.

And yet, given five minutes and a napkin, you could land within a factor of three of the true answer. Not by remembering anything, but by reasoning — breaking one impossible question into a handful of merely difficult ones, guessing at each, and multiplying. This is estimation, and it is the closest thing to a superpower that arithmetic offers.

The core move: decompose

The reason “how many gallons of paint?” feels unanswerable is that it is one enormous question. The reason it becomes answerable is that it is secretly four small ones:

• How many houses are there in America?
• How much wall does one house have?
• How much wall does a gallon of paint cover?
• How many coats?

Every one of those is a question you have some grip on. You know roughly how many people live in the United States. You have stood next to a house. You may have bought paint, and if you haven't, you can reason that a gallon is a decent-sized jug and walls are big, so it probably covers a room-ish amount rather than a stamp or a stadium.

None of your four answers will be right. That turns out not to matter nearly as much as you'd think, for reasons we'll come to.

The four steps

Doing it for real

Let's actually paint America. Roughly 350 million people live in the United States. Households average somewhere around 2.5 people, so call it 140 million households. Not all are detached houses — a lot of people live in apartments — so maybe 90 million of those are individual houses. Round to 100 million, because we're not being precise, we're being close.

Now, one house. Picture a modest two-storey home: maybe 40 feet across, 30 deep, 20 tall. The four walls come to roughly (40 + 30 + 40 + 30) × 20 = 2,800 square feet of exterior. Knock some off for windows and doors, add some back for gables and trim, call it 2,500 square feet.

A gallon of exterior paint covers something like 350 square feet — that number is printed on the can, and if you've never read one, "a few hundred square feet" gets you the same answer. So one house needs about 7 gallons a coat, and houses get two coats, so 15 gallons.

About 1.5 billion gallons. Is that right? The honest answer is: it's the right size. The US paint industry ships on the order of a billion gallons a year across all uses, and houses get repainted every decade or so rather than annually, so our figure — a full repaint of every house in the country at once — sits sensibly above one year of production. The estimate passes the sniff test, which is all it was ever meant to do.

You now know something you did not know ten minutes ago, and you learned it without looking anything up.

Why the errors don't pile up

The objection everyone raises: surely four sloppy guesses multiplied together produce one catastrophically sloppy answer?

Remarkably, no. The reason is that your errors go in both directions. You overestimated the number of houses; you underestimated the wall area. Multiplying a number that's 1.5× too big by one that's 1.5× too small returns you to exactly where you started. Mathematically, multiplying estimates is equivalent to adding their logarithms — and independent errors added together behave like a random walk, drifting rather than compounding.

The more factors you break a problem into, the more chances your mistakes get to cancel each other out. Estimation gets more reliable as it gets more elaborate — the opposite of what intuition suggests.

The catch is the word independent. If you're systematically optimistic — if every single guess leans the same direction because you want a particular answer — the errors stack instead of cancelling. Which is a decent argument for making your guesses before you decide what you're hoping to prove.

Where this actually matters

Estimation is not a parlor trick. It is the skill that lets you hear a claim and know, within seconds, whether it deserves your attention.

A politician says a program will cost $40 billion. Is that a lot? Divide by 350 million Americans: about $115 each. Now you have an opinion instead of a reaction. A friend says a startup is worth $2 billion. What would that require in revenue? At a typical multiple, hundreds of millions a year — so how many customers, at what price? Suddenly you can ask a useful question.

Engineers use this before they build. Doctors use it before they prescribe. Anyone who has ever caught an obviously-wrong number in a spreadsheet used it without naming it. The point of practising is that the estimate arrives fast enough to be useful in the moment, rather than an hour later on a napkin.

Getting better at it

Three habits, in ascending order of usefulness.

Collect anchor numbers. A handful of memorised quantities — world population, the population of your country, the speed of sound, the number of seconds in a year — do enormous work, because most estimates eventually reduce to one of them.

Estimate before you look up. Whenever you're about to search for a number, guess first. It costs three seconds and turns an idle lookup into a calibration exercise. Over months, this quietly rebuilds your intuition.

Notice which way you were wrong. Most people are consistently biased in one direction on certain kinds of quantity — almost everyone underestimates very large numbers, because human intuition has no native feel for the gap between a million and a billion. Knowing your own tilt lets you correct for it.