Validating Search Console Bulk Export Partitions and Retries
When Search Console sends performance data to BigQuery, it writes each day's data into date-based partitions, which keep each date's rows together.…
In Google Search Console’s Search results Performance report, the chart shows performance over time, while the table breaks performance down by a selected dimension, such as query or page. An export downloads report data for analysis. Their totals can differ because Google does not always count, filter, or return the same records in each view.
The first check is whether the views use the same property, date range, search type, metrics, and filters. Then check how Google grouped the data and whether the table or export omits rows. A difference alone does not show that one view is wrong; the number to use depends on whether you need an overall property total or a breakdown of reported rows.
The chart is generally useful for an overall total for the selected property and period. The table answers a different question: how that performance is distributed across a chosen dimension. Google’s aggregation rules vary by dimension. Query, country, device, and date rows use property-level aggregation, while page and Search appearance rows use page-level aggregation.
At property level, Google counts a property once for a search result even if several URLs from that property appear. At page level, Google counts each unique URL separately. A page-grouped table can therefore have more impressions than the property-level chart. For example, if three URLs from one property appear for the same search, the chart can count one property impression while a page-grouped table counts an impression for each URL. Google explains these differences in its data aggregation guidance.
The same distinction affects click-through rate (CTR), which is clicks divided by impressions, and average position. Do not add CTR or average-position values across rows or take a simple average of them. To calculate a combined CTR from rows, divide the rows’ total clicks by their total impressions. Use the chart when the question concerns the overall property total; use page rows when the question concerns individual URLs.
An export can contain chart data and dimension rows, but the dimension rows may not include every result. Google says Search Console tables and report exports can be limited to 1,000 representative rows, while report totals include truncated data. Check which part of the export you are summing: a total shown in the export can cover more data than the exported dimension rows. The export documentation also says values displayed as ~ or - are exported as zero, which can affect spreadsheet calculations.
Search Console hides some low-volume queries to protect searcher privacy. Those anonymized queries are generally included in chart totals but omitted from query rows. If you filter the report by query, the filtered result also excludes anonymized queries. As a result, visible query rows can sum to less than the chart, and two complementary query filters may not add back to the unfiltered total. Google’s discrepancy guidance describes this behavior.
Tables also have row limits. When a report contains more rows than Search Console returns, rare or long-tail rows can be missing from the displayed table even though their activity contributes to the total. The same principle applies when analyzing API results: the Search Analytics API allows up to 25,000 rows per request, but Google’s data retrieval guidance describes a limit of 50,000 rows per day per search type. Those limits restrict returned detail; they do not guarantee that the returned rows add up to the chart.
Filters and freshness can also affect a comparison. Google documents that page or Search appearance filters can change how chart and table data are aggregated. The newest chart data can be preliminary and later change as Google processes it. The API’s dataState setting controls whether a request includes fresh data; without it, the API returns finalized data. Compare the same dates and data state before treating a difference as a reporting problem.
Use the number that matches the question and label its scope clearly:
Before comparing numbers, record the property and date range for both views. For each, note the search type, selected metrics, filters, and any comparison setting. Then check whether the table groups by page or Search appearance while the chart reports property-level totals.
Next, compare the chart total with the sum of the table rows, but treat the sum as a row total rather than a replacement for the chart. A shortfall in query rows is consistent with privacy filtering; a shortfall in a large table or export can reflect row limits. For exports, confirm whether you are summing chart data or dimension rows, and check how the spreadsheet handles ~ and -.
For an API comparison, also record the dimensions, filters, aggregation type, pagination, and dataState. If the difference remains after those settings match, check Google’s Search Console data anomalies page for an incident that applies to the report and dates in question.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.com