I started with upload rate per minute, which felt right, then tried to back into total volume by estimating how long YouTube has existed and how upload rates have grown over time.
Break the problem into supply (video uploads) and demand (views/watch time) sides, then estimate the number of videos using a top-down or bottom-up approach. Use assumptions about YouTube's user base, upload frequency, and video lifespan to derive a reasonable range. Clearly state assumptions and validate with sanity checks.
Pro tip: Anchor your estimate to a known metric like daily uploads (e.g., 500 hours of video uploaded per minute) and convert to video count using average video length, then adjust for historical accumulation and deletions. This shows you can leverage public data and think about stock vs. flow.
Define what 'videos currently on YouTube' means: public, unlisted, private, live streams, shorts, etc. Ask if the interviewer wants a rough order of magnitude or a precise number.
Decide between top-down (using total watch time or upload rate) and bottom-up (using number of creators and upload frequency). For a PM role, a top-down approach using known metrics like '500 hours uploaded per minute' is often efficient.
List key assumptions: average video length (e.g., 7 minutes), percentage of uploads that are public and remain on the platform, growth rate over time, and deletion rate. Use round numbers for easy calculation.
Convert upload rate to videos per minute, then extrapolate over YouTube's lifetime (since 2005). Adjust for non-public videos and deletions to estimate current stock. For example, 500 hours/min = 30,000 minutes/min; at 7 min/video, ~4,286 videos/min; over ~18 years, that's ~40 billion, but adjust for growth and deletions.
Compare your estimate to known figures (e.g., YouTube reported 800 million videos in 2019, but that's likely outdated). Check if the number seems plausible given daily uploads and total watch time. Adjust assumptions if needed.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.