Start by reading the shape of the drop. A sudden fall on a specific date points to something that changed: a migration, a release, a robots or canonical error, or a manual action. A gradual decline across months points to competition, changing search results, or content that has aged. Those two shapes have almost nothing in common, and treating a slow slide as a technical fault wastes a quarter.
First, confirm it is real
Before diagnosing anything, rule out the boring explanations. A surprising share of reported drops are measurement artifacts.
- Compare like with like. Four weeks against five weeks produces a drop that is not there.
- Check for seasonality by comparing with the same period last year rather than last month.
- Check whether the tracking code is still firing. A release that removed a tag looks exactly like a traffic collapse.
- Check whether the drop appears in Search Console clicks as well as in analytics. If only one shows it, the problem is measurement.
- Check whether a consent banner changed. Consent changes move analytics numbers without moving a single visitor.
Then read the shape
| Shape | Usual cause | Where to look |
|---|---|---|
| Cliff edge on one date | Something you or your team changed | Deployment history, robots.txt, canonicals, redirects, hosting |
| Cliff edge with no release | A core update, or a manual action | Update timelines, and the Manual Actions report |
| Slow slide over months | Competition, ageing content, changing results pages | Who ranks now against who ranked then |
| One section only | Something specific to that template or directory | That template, its internal links, its indexation |
| Impressions flat, clicks down | The results page changed shape around you | What now appears above the first organic result |
That last row is worth dwelling on. Stable rankings with falling clicks usually means something new is sitting above you, taking the click before anyone scrolls. Your position did not move. The page around it did.
The order of operations
- Check the Manual Actions report. Thirty seconds, and it rules out the worst case immediately.
- Check indexation. A sharp fall in indexed pages explains a sharp fall in traffic and points straight at a technical cause.
- Line the drop date up against your deployment history. If they match, you have your answer and can stop.
- Check robots.txt and a sample of canonical tags. A staging robots file shipped to production is a classic, and it is invisible until you look.
- Segment by page and by query. A site-wide drop and a drop concentrated in one directory are different problems.
- Only then consider an algorithm update, and check the date against known update timelines rather than assuming.
What not to do
- Rewriting large amounts of content in the first week, before the cause is known
- Disavowing links reflexively, which more often causes harm than fixes anything
- Making several large changes at once, which makes the recovery impossible to attribute
- Assuming an update, because it is the explanation that requires nobody to have made a mistake
Not every drop has a fixable cause. Sometimes a competitor got genuinely better, or the search results for your terms now answer the question without a click. Those need a different response from a technical repair, and identifying which you are facing is most of the work.