Diagnosing a traffic drop: a decision tree

A traffic drop triggers the worst instinct in site ownership: change things until the fear goes away. Rewrites, redesigns, panic-deletions — applied before anyone establishes what actually happened. The productive response is a diagnosis sequence, and it's short: verify, date, scope, then match the pattern. Work it in order; each step eliminates half the remaining suspects.

Step 1: verify the drop is real

A meaningful share of 'traffic drops' are measurement drops. Check Search Console clicks against your analytics for the same date range. If analytics fell but Search Console clicks didn't, your traffic is fine and your tracking broke — a tag removed in a site update, a consent banner change, a plugin conflict. Fix the measurement and stand down. Only when both sources agree do you have a real drop worth diagnosing.

Step 2: date it precisely

Zoom the Search Console performance graph until you can name the day the decline started. Cliff or slope matters enormously: a cliff (normal Tuesday, 40% down Wednesday) points to an event — a site change, a technical break, an algorithm update rolling out. A slope over weeks or months points to erosion — competitors, decay, shifting search behavior. Then check the date against two lists: Google's published update timeline, and your own deploy/change history. The overlap is usually the answer.

Step 3: scope it — everything, or some things?

In Search Console, compare the drop period against the prior period, split by page and then by query. Three shapes:

• Everything fell proportionally — sitewide event: technical problem, algorithm update, or a sitewide change you shipped.
• A few pages fell hard, rest fine — page-level: those pages got outranked, decayed, or a SERP feature landed on their queries. This is the underperforming-post workflow, not an emergency.
• Clicks fell, impressions held or rose — you're still ranking; the results page changed around you. Increasingly, that's an AI Overview arriving on your informational queries, or new ads/features pushing you visually down.

Step 4: match the fingerprint

With date and scope in hand, one of six patterns almost always fits:

• Tracking break — analytics down, Search Console fine. Fix the tag; apologize to nobody.
• Technical self-inflicted — cliff, sitewide, matches a deploy: an accidental noindex, broken redirects after a migration, robots.txt mishap, or the site was simply down. Check Search Console's indexing report; the not-showing-up diagnostic covers the usual suspects. Fastest to fix, most common on small sites.
• Algorithm update — cliff, sitewide or content-wide, matches Google's update calendar, competitors in your niche wobbling too. No same-week fix exists. Response: keep publishing, improve the content the update seemed to dislike, and wait for the next cycle; recovery happens on Google's calendar.
• Outranked — slope, page-scoped, position drifting down while someone else's drifts up. Search your queries, read who beat you, refresh accordingly.
• Seasonality — the drop matches last year's same months. Verify with year-over-year comparison in Search Console before diagnosing anything else; businesses rediscover their own seasonality with alarming regularity.
• SERP changed around you — impressions steady, clicks down, position steady. AI Overviews, featured snippets, more ads. You didn't get worse; the page of results did. Response: target the queries that still click — commercial and comparison intent — and make your snippets earn the clicks that remain.

Date plus scope identifies almost every traffic drop: cliffs are events, slopes are erosion; sitewide is technical or algorithmic, page-level is competitive.

What not to do while diagnosing

• Don't ship sweeping changes during an algorithm update rollout — you'll be optimizing against a moving target and polluting your own evidence.
• Don't delete content in panic; deletion is the last treatment, not the first.
• Don't buy emergency SEO services sold on fear — a drop is when the red-flag vendors find you, and 'we'll recover your rankings' is their best-converting pitch.
• Don't judge the recovery daily. Weekly is the honest cadence; the signals that matter move on weekly timescales.

The prewritten incident file

The whole diagnosis takes an hour if you have two artifacts ready before anything drops: a change log (every deploy, migration, plugin, and content change with dates — a text file suffices) and your baseline numbers (the monthly five-number row you're already keeping). Drops diagnosed against records take an hour; drops diagnosed against memory take a week and end in superstition. The difference is a text file you can start today.