Estimation question

Estimate the daily bandwidth consumed by YouTube globally.

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

Technical estimation of network bandwidth combining user base, watch time, and video quality assumptions.

Anchor on YouTube's global user base, roughly 2.5 billion monthly active users, and estimate daily active users at a fraction of that, say 1 billion.
Estimate average daily watch time per active user, say 40 minutes, based on widely reported YouTube engagement figures.
Calculate total daily watch minutes: 1 billion users times 40 minutes gives 40 billion minutes watched per day.
Assume an average bitrate across all quality levels (mixing mobile, SD, and HD viewers), say 3 Mbps blended average.
Convert minutes to seconds and multiply by bitrate to calculate total daily data transferred, landing in the tens of exabytes range.
Divide by seconds in a day to estimate an average bandwidth figure, then note that peak hour bandwidth would be several times the daily average.

Breaks the problem into DAU, watch time per user, and bitrate as three separate, clearly stated assumptions.
Uses a blended bitrate reflecting the real world mix of mobile, SD, and HD streaming, rather than assuming uniform HD quality.
Shows clear unit conversions from minutes to seconds and from bits to bytes, avoiding a common order of magnitude error.
Distinguishes average daily bandwidth from peak hour bandwidth, which is materially higher and matters more for infrastructure planning.

Assuming all video streams at HD quality, significantly overstating real world blended bandwidth.
Making unit conversion errors between bits, bytes, minutes, and seconds that throw off the final order of magnitude.
Not distinguishing average bandwidth from peak bandwidth, which matters far more for actual infrastructure decisions.

Chapters of the AI PM course that teach what this question tests.