Estimation questions in job interviews

“How many windows are there in New York?” is not a trick question, and there is no right answer. There is, however, a right way to answer — and it is entirely learnable.

Zoggle · a guide for people who like numbers

Management consultancies have asked these for decades. Tech companies picked them up, largely abandoned them after concluding that brainteasers predicted nothing, and then quietly kept the good ones — because a well-run estimation question does measure something real: whether a candidate can impose structure on an ambiguous problem while someone watches.

The mistake nearly every candidate makes is treating it as a knowledge test and panicking about not knowing the answer. Nobody knows the answer. The interviewer usually doesn’t either.

What is actually being scored

Four things, roughly in this order of weight:

• Structure. Do you decompose the problem before diving at it, and is the decomposition sound?
• Communication. Can the interviewer follow your reasoning without asking you to repeat yourself?
• Judgement. Are your assumptions defensible, and do you flag which ones you’re least sure about?
• Composure. When you’re challenged mid-answer, do you adapt or unravel?

The final number matters far less than any of these. An answer that is 5× off but arrived at cleanly beats a lucky guess with no visible reasoning, every time.

The structure that works

A worked example, out loud

Windows in New York City. Roughly 8 million people. At about 2.5 per household, that’s 3.2 million households — call it 3 million. New York is dense, so most are apartments; say an average apartment has 4 windows. That’s 12 million residential windows.

Commercial next. Office space is harder, so approach it from workers: maybe 4 million people work in the city, at perhaps 200 square feet each, giving 800 million square feet of office space. A window per 100 square feet of floor is a reasonable ratio for a glassy city, so about 8 million commercial windows.

Then the check: does that pass? Twenty million windows across 8 million people is about 2.5 windows per resident, which feels low for a city that also contains offices and shops — suggesting my apartment figure of 4 windows is conservative, and the true answer is probably somewhat higher. Saying that last part out loud is worth more than the calculation.

Handling the pushback

Good interviewers challenge an assumption mid-answer, deliberately. “Four windows per apartment seems low to me.” They are not correcting you; they are testing what happens next.

The wrong responses are caving instantly and defending rigidly. The right one engages: “That’s fair — studios might have two, larger apartments six or more. If I use six instead of four, residential goes from 12 to 18 million and the total moves to about 26 million. It doesn’t change the order of magnitude, which is reassuring.”

That answer demonstrates something valuable: you know which of your inputs actually matter. Recognising that an assumption barely moves the result is a genuinely senior instinct.

Preparing without memorising

You cannot study the specific question, and trying to memorise worked examples is how people end up reciting a rehearsed answer to a different question. What transfers is the habit.

Learn a handful of anchor numbers — populations, household sizes, rough salaries, the size of a room. Practise on real questions until narrating your reasoning while calculating stops feeling like patting your head and rubbing your stomach. That dual-tasking is the actual difficulty, and it only yields to repetition.