Search clicks dropped
Where it sits, not what caused it. Two exports, before and after, read side by side until the change is located in the rows that carry it.
- Price
- On request, paid up front, tax inclusive
- Turnaround
- 2 working days from when we have everything we need
- You send
- Two untouched Search Console downloads, as the zip files arrived, and a screenshot of each export screen
- You get
- One written read of where the returned rows changed
- Calls
- None. Everything in writing.
- Refund
- Full, on request, and you keep the review
What we read
Both exports as Search Console wrote them, each sent as the zip file it arrived in: queries, pages, countries, devices, search appearance and the daily chart. Files lifted out of a download are refused, because a folder somebody filled can hold two windows and look exactly like one export. The periods must not overlap, and must match on property, search type, filters, number of days and weekday composition.
The read shows where the returned rows changed by page, search, country, device and date. It works out a share of the overall change only where both sides count the same thing the same way; where they do not, the difference is given at its size instead. Search appearance is read but not summed into a movement, because its rows can overlap.
What the exports leave out
No file in a download names the property it came from, and none marks a day Search Console had not finished counting. That is why a screenshot of each export screen is part of what you send: the property, the date range, the search type, that no other filter is set, and a chart with no preliminary days. Without it those two facts are reported as your statement rather than as anything we checked.
Country, device and date rows are counted across the whole property; page rows are counted per address. Returned search rows can leave data out, and one impression can count under more than one kind of search result. A gap between those tables is stated at its size without assigning it one cause.
A query table at the interface's row limit is a returned sample of an unknown whole. A row present on one side only may have disappeared, appeared, or been crowded out by other rows, so the review keeps those states separate rather than turning absence into zero.
What it does not claim
A cause. An export records what was shown and clicked, not why. Where the export cannot distinguish among plausible explanations, the review says so, and names additional evidence only when it would actually resolve the difference.
Questions
Why two exports rather than one comparison file?
Two untouched downloads with matching filters and equal periods are something we can verify line by line. A comparison file is not, yet.
My analytics shows a drop too. Same thing?
Not necessarily. Search Console counts clicks from Google Search, analytics counts visits from everywhere, and the two can move for different reasons.
What if nothing in the export explains it?
The review says so and names what the export cannot see. The refund stands.
Request the review
Send the site and enough context to check whether the review fits. Do not send account access or client material through this form.