Root-Cause Analysis of a 20% Single-Day Decline in YouTube Engagement Time

Time spent on YouTube fell 20% compared with the previous day. Explain what you would do: how you would work out why it dropped, and how you would respond once you know the cause. ```hint Let the speed guide you A fifth of a day's usage vanishing overnight is far faster than habits normally change. Let the size and suddenness of the move decide which explanations you check first. ``` ### Clarifying Questions - How is time spent measured here: total minutes across all users, the average per daily active user, or the average per session? Does it include Shorts, the TV apps, YouTube Music and YouTube Kids? - Which two days are being compared (for example, a weekday against a holiday or the day after a major live event), and has the metric recovered since? - Did the number of daily active users fall too, or only the time each active user spends? - Is the decline global, or concentrated in particular countries, platforms or app versions? ### What a Strong Answer Covers - Ruling out logging, pipeline and definition problems before treating the drop as a real change in behavior. - Decomposing time spent into users, sessions per user and time per session to see which factor moved. - Segmenting by region, platform, app version, content format, traffic source and viewer versus creator to localize the drop. - Checking internal causes (releases, experiments, outages, ad or recommendation changes) before external ones, each with the evidence that would confirm it. - A response matched to the confirmed cause, plus how to detect a repeat sooner. ### Follow-up Questions - The drop is confined to the newest Android app version, and daily active users are flat. What do you do in the next hour? - Time spent fell 20%, yet ad revenue for the day barely changed. How could that happen? - After every internal cause is ruled out, the drop is limited to one country. What would you look into next?