Estimation question

Estimate the number of emails that the city of San Francisco receives.

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Whether you can decompose an unfamiliar estimation problem into clean, stated assumptions.

Clarify scope: emails received by people and organizations located in San Francisco, per day or per year.
Estimate the population: roughly 800,000 residents plus commuters and businesses receiving mail there.
Split into segments: personal emails per person per day, and business or transactional emails per worker per day, since volumes differ a lot.
State assumptions: roughly 20 personal emails per person per day, and 100 plus emails per office worker per day including automated mail.
Multiply each segment by its population and sum, then convert to an annual figure if the question calls for it.
Sanity check the total against known scale, like global daily email volume in the hundreds of billions, to confirm the city's share is reasonable.

Splits personal versus business email volume, since office workers receive far more email than the average resident.
States clear numeric assumptions at each step instead of asserting a final number.
Accounts for commuters and businesses, not just registered residents, since a city's email volume includes both.
Sanity checks the result against a rough sense of global email volume to catch an order-of-magnitude mistake.

Using one flat per-person email estimate that ignores the much higher volume from business accounts.
Skipping a sanity check, leaving an estimate that could be off by 10x unnoticed.

How would this estimate change if you included automated and marketing emails separately?
Which assumption in your estimate are you least confident about?