How we date a traffic drop to within four days

When a site loses traffic, the first meeting usually produces a list of theories: a Google update, a competitor, the new theme, the hosting move. Every theory sounds plausible. Without dates, there is no way to choose between them, and the budget goes to whichever one the loudest person believes.

So before we change anything, we date the decline. On a recent engagement we traced the main cause of a collapse to one site update, dated to within four days and proven five different ways.

1. Get enough data to see the shape

The Search Console interface shows a few months and a sample of pages. That is not enough. We pulled 16 months of search data, 7.35 million rows, by page and by search phrase. With that, a decline stops being one falling line and becomes several smaller drops, each starting on its own day.

2. Split the line until it explains itself

• By page type. Did articles fall while category pages held? That points at templates or content, not at the whole domain.
• By impressions versus clicks. Fewer impressions means Google is showing you less. Same impressions but fewer clicks means you are still visible, but losing on titles or features.
• By page count. If the number of pages Google shows falls, something is removing or hiding them.

3. Put every change on the same timeline

Code releases, plugin updates, settings changes, hosting work, editorial decisions, scheduled jobs. Most teams have these scattered across tools and memories. Put on one timeline, a drop that starts the same week as a release stops being a coincidence worth ignoring.

4. Try to prove yourself wrong

This is the step most audits skip. Our main finding faced nine attempts to disprove it before it went into a report. Did the drop start before the release? Did unaffected pages fall too? Does a Google update fit the dates better? An explanation that survives that is worth spending money on. One that does not would have sent the budget the wrong way.

Why it pays

Dating the decline told us which fixes mattered and in which order. In the first month, weekly visits from Google went from 602 to 1,612. That was not luck: the work went to the causes, not to the theories.