Root-Cause Analysis for a 50% Drop in Dropbox File Uploads

You are a product manager at Dropbox, and you learn that uploads are down 50%. What would you do? Walk through how you would confirm the drop, track down its cause and respond. ```hint Narrow it before you explain it Before guessing at causes, think about which few cuts of the data would tell you fastest whether the measurement is broken, the product is broken, or people really changed what they do. ``` ### Clarifying Questions - Does "uploads" mean upload attempts, completed uploads, the number of files or the bytes uploaded? - Compared with what baseline (yesterday, the same day last week, last month), and did the drop appear all at once or build up over time? - Is the drop spread across every platform (web, desktop app, mobile apps), region and account type, or concentrated in some of them? - Were there releases, experiments, pricing changes or outages around the time the drop began? ### What a Strong Answer Covers - A precise definition of the metric and its baseline, and a check of whether the drop is real or a logging or pipeline artifact - A structured set of hypotheses (internal changes, technical failures, external platform changes, user behavior and competition) with a sensible order for testing them - Segmentation by timing, platform and client version, region, account type, upload source and funnel stage to isolate the cause - Fast mitigation and communication kept separate from root-cause analysis, with recovery tracked - Prevention: earlier detection, safer releases and a post-incident review ### Follow-up Questions - Suppose the drop is only on the web client, only for users in North America, and built up over about two days. What do you check first? - Upload attempts are flat but completed uploads have halved. What does that tell you, and whom do you bring in? - How would you set up monitoring so that a drop like this is caught within the hour instead of found in a report?