Diagnosing a Sudden Drop of Website Traffic to Zero

You are a product manager, and you learn that traffic to your company's website has suddenly dropped by 100%: the visit count that looked normal shortly before now reads zero. What would you do? Walk through how you would respond, how you would find the cause and how you would make sure it does not happen again. ```hint Is the zero real? Demand rarely falls to exactly zero at once. Before looking for reasons people stopped coming, ask which kinds of failure would produce a perfectly flat line, and how you would tell them apart. ``` ### Clarifying Questions - How did the drop unfold: a cliff at one moment, or a decline over several hours? Did it coincide with a deploy, a configuration change or a renewal date? - Where does the zero come from: only the web analytics dashboard, or also server logs, CDN request counts and server-side events such as sign-ups and orders? - Does the drop cover every page, region, device, browser and channel, or does any slice still show visits? - Do the mobile apps and other non-web surfaces still show normal usage? ### What a Strong Answer Covers - Treating a total drop as a live incident: urgency, an owner and communication alongside the investigation. - Verifying that the drop is real by comparing the analytics with independent signals before chasing user-behavior explanations. - A diagnosis that separates measurement failures, site availability and access-path problems, checked in an order that fits a sudden, complete drop. - Using the exact time of the cliff and segment breakdowns to narrow the cause, and recognizing which candidate causes cannot explain a drop to zero. - Recovery and prevention: mitigation, confirming that traffic fully returns, and monitoring and redundancy so it does not repeat. ### Follow-up Questions - Logs show one data center went down, yet a global site should have failed over to another. What else would you investigate? - Suppose the interviewer reveals that visits fell to zero only for one browser while every other browser looks normal. What could cause that, and how would you confirm it? - After the fix, what monitoring would you add so the team learns about the next outage from an alert instead of from a dashboard?