Product Sense · medium · Reported at Lyft
Estimate the number of ride requests in a city per day.
Asked in Data Analyst interviews.
Short answer
Market sizing: population times adoption times trips per day, with every assumption stated.
How to answer it
What they are testing
Not the answer. Whether you can decompose an unknown into things you do know, state your assumptions, and sanity-check the result. Arithmetic under pressure is a small part; structure is most of it.
The method
1. Anchor start from a population you actually know
2. Decompose break the unknown into rate x frequency
3. Estimate put a number on each factor and SAY why
4. Multiply keep the arithmetic round and easy to follow
5. Sanity check the order of magnitude against something knownA worked estimate
City population 5,000,000
Adults, roughly 75% 3,750,000
Use ride-hailing at all, say 25% 940,000
Of those, ride on a given day, 15% 140,000
Rides per riding day, average 1.6 225,000
Round: roughly 200,000 to 250,000 ride requests a dayEvery line is a number you can defend or be corrected on, which is the point. An interviewer who disagrees with the 25% can change it and watch the answer move, and that conversation is what they are actually buying.
Sanity checks
- Rides per person per year: 225,000 times 365 divided by 5,000,000 is about 16. Plausible for a large city.
- Drivers implied: 225,000 rides at maybe 15 rides per driver per day is 15,000 drivers. Does that sound like the city?
- If a number implies something absurd, say so and revise. Catching your own error scores better than defending it.
Key points
- Anchor on a population figure, not on a guess about the whole answer.
- State each assumption as you make it, so it can be challenged.
- Keep the numbers round. Precision you do not have is a distraction.
- Sanity-check with a second, independent route to the same quantity.
- Give a range at the end, not a false-precision single number.
FAQ
- What if I do not know the population? Ask, or anchor on something you do know and say so. Made-up precision is worse than an assumption.
- Should I account for time of day? Only if asked. Mention that demand is peaky and move on.
- How exact should it be? Order of magnitude. Being within a factor of two is a good outcome.
The 30-second answer
I would build it up from a population I know rather than guessing the answer. Say a city of five million: about three quarters are adults, maybe a quarter of those use ride-hailing at all, and perhaps fifteen percent of those take a ride on any given day, at around one and a half rides each. That lands near two hundred thousand requests a day.
I would say every assumption out loud as I make it, because the interviewer can then correct one and we watch the answer move, which is the real conversation.
Then I would sanity-check it a second way. Two hundred thousand rides a day across five million people is about sixteen rides per person per year, which is plausible, and it implies roughly fifteen thousand drivers. If either number looked absurd I would go back and revise.
In the interview
- What they are checking: a clean decomposition, explicit assumptions, and a sanity check at the end.
- Common mistake: precision theatre. Quoting 147,320 from assumptions that are each plus or minus 50%.