Diagnosing a June Surge in Video Uploads and Validating the Root Cause

You are the product manager for a platform where people upload their own videos, and the dashboards show that the number of user-uploaded videos spiked in June. Explain what you think caused the spike and how you would test that explanation. ### Clarifying Questions - How large was the spike, and did it arrive as a sudden jump on particular days or as a steady climb through the month? - Does it appear on every surface (web, iOS, Android) and in every region, or is it concentrated somewhere? - Did more people upload, or did the same people upload more videos each, and were the uploaders new accounts or existing ones? - Did uploads rise like this in June of previous years? ### Part 1 — Generate and rank the possible causes List the explanations you would consider for the jump, and put them in the order you would investigate them, weighing how likely each one is against how cheap it is to check. ```hint Question the number before the behavior Ask first whether the count itself could be misleading. Then separate changes your own company made from changes in the world around the platform. ``` #### What This Part Should Cover - A check that the spike is real: logging and metric definitions, duplicate uploads, and automated or spam uploads. - Internal causes such as launches, fixes to the upload flow, promotions or creator incentives. - External causes such as seasonality, one-off events, problems at competing platforms, and changes in devices or connectivity. - A breakdown of the metric (number of uploaders times uploads per uploader, cut by segment) that narrows the list. ### Part 2 — Test the leading explanations For your top explanations, describe the evidence that would confirm or rule out each one, and how you would establish that it actually caused the spike rather than merely coinciding with it. ```hint Find the cut that separates them Each explanation leaves a different fingerprint: which users uploaded, the day it started, which app version they were on, and what happened in the weeks after. Pick the view of the data where your top candidates disagree. ``` #### What This Part Should Cover - A distinct, checkable prediction for each hypothesis. - Methods that fit whether the cause can be randomized: an experiment or holdback where it can, otherwise comparisons across versions, regions, cohorts or time. - A check that the chosen explanation accounts for the full size of the spike, not just a part of it. ### What a Strong Answer Covers - A structured, prioritized investigation rather than an unordered list of possible causes. - Data validity confirmed before behavioral explanations are entertained. - Tests that move from correlation toward causation wherever possible. - A view on whether the spike is healthy (views per video, moderation load, storage cost, whether new uploaders stay) and what the team should do next. ### Follow-up Questions - The cause turns out to be an outage at a competing video platform. How would you keep the creators who arrived during it? - You discover a release that fixed a bug which had been making some uploads fail. How would you estimate how many uploads were lost before the fix, and what would you change so it is caught sooner next time? - Uploads rose, but views per video fell. What does that tell you, and does it change your recommendation?