Traffic dropped suddenly: how to diagnose it without panicking
Traffic fell and you do not know whether you were penalised, deindexed, or are looking at a chart artefact.
Look at the shape before the cause
A gradual decline over weeks is a demotion: pages still indexed, ranking worse. A drop to zero or near-zero within a day or two is deindexing: the pages are not in the index at all. The two share almost no remedies.
Split by channel immediately
A blended total hides a dead channel. One engine on this site went to zero for a month while the other had its best month ever, and the combined chart barely moved. Review channels separately, weekly.
Escalate in order of cost
Console messages first. Then resubmit and request reindexing. Then infrastructure: robots.txt, DNS, certificates, whether a deploy changed a header. Content last. Most sudden drops are resolved before you reach content, and starting at content means you will never know what fixed it.
Separate the engine from the query set
Before anything else, split the drop by search engine and by branded against non-branded. A drop that is entirely Bing is probably Bing, and this site lost 31 straight days to exactly that. A drop that is entirely brand is a demand problem, not a search problem. A drop across non-brand queries on Google is the only one that means what you are afraid it means.
The drop that was not a drop
Check whether you are comparing against a spike. One viral week, one seasonal peak, or one post that got shared will make the following month look like a collapse. Set the comparison to the same period a year earlier, or to the trailing 90 days, before you conclude anything. More reported drops are baseline errors than are algorithm updates.
Measured, not asserted
The largest single anomaly in this dataset was a 31-day channel outage caused by nothing the site did, and resolved by asking for a recrawl. The month it happened, the other channel grew 61%.
Free tool for this: Noise or Signal. No account, nothing uploaded.
Where this goes deeper
Every number on this page comes from one complete dataset: one product taken from zero to 100K+ users on search alone, with nothing spent on advertising. The full argument is Chapter 22 of the book. Five chapters are free to read.