Start With Two Comparable Search Performance Periods
Choose one earlier period as a baseline and one later period as the comparison. For regular monitoring, two consecutive 28-day periods have the same number of weekdays. Comparing the same season in different years may be more useful for highly seasonal websites. Avoid comparing raw clicks from periods of unequal length.
Keep the property, search type, country, device, and any query or page filters identical. Do not compare a mobile-only export with an all-device export or Web Search with Image Search. Record each date range separately.
Export the Right CSV Files From Google Search Console
- Open your verified property in Google Search Console and select Performance > Search results.
- Set the earlier date range, search type, and any other filters you intend to keep.
- Open Queries if you want search-term changes, or Pages if you want URL changes.
- Use Export > Download CSV. If your download contains multiple CSV files, select the file for the dimension you chose.
- Repeat with the later date range, keeping every other setting the same.
Use two separate, same-dimension exports for a straightforward earlier-versus-later analysis. Search Console also has a built-in Compare mode, but a comparison-mode export can have a different column structure; do not assume a CSV tool supports that structure unless its interface says so. Keep the original files for auditing.
Why Queries.csv and Pages.csv Answer Different Questions
A Queries export groups performance by search term. It helps identify growing questions, declining topics, and searches with impressions but few clicks. A Pages export groups by URL. It helps reveal individual articles, product pages, or directories that need investigation. Match Query to Query or Page to Page; never combine those dimensions as though their rows represented the same entities.
Google commonly attributes page-level Performance data to its selected canonical URL. A change in canonical selection, migration, or URL structure can therefore affect the visible page rows. Check URL Inspection and redirects for missing or replaced pages. For a second layer of analysis, group comparable query topics with the Keyword List Grouper, or assess competing URLs with the Keyword Cannibalization Analyzer.
Read Clicks, Impressions, CTR, and Position Together
For one query or page, calculate click difference = later clicks minus earlier clicks. Percentage click change is that difference divided by earlier clicks, multiplied by 100. If earlier clicks are zero, conventional percentage growth is undefined; report the absolute increase instead. Calculate CTR from clicks divided by impressions. For grouped rows, calculate CTR using summed clicks and summed impressions, not the arithmetic mean of row CTRs. Likewise, an aggregate average position should be weighted by impressions rather than averaging row positions equally.
A Worked CSV Comparison With Two Different SEO Signals
The following figures are illustrative examples, not measurements from a live website. Both examples compare equal-length periods using the same report dimension and filters.
Example A gained 30 clicks (+25%) and 1,000 impressions (+25%). CTR did not improve; the page simply received more recorded exposure while its average position improved numerically. Example B lost 20 clicks (-25%) despite gaining impressions. Its CTR fell by 1.5 percentage points, and its average position worsened. Investigate its queries, search results, and page relevance.
Subtracting positions works in the opposite intuitive direction from clicks: later position minus earlier position is negative when the average position improves. A change from 8.4 to 7.6 is an improvement of 0.8 positions, not an 0.8-position decline.
Separate New Export Rows From Genuine New Rankings
If a query appears only in the later CSV, label it present only in the later export. If it appears only in the earlier CSV, label it present only in the earlier export. Neither proves that Google just discovered or removed a ranking. It may have fallen below an export limit, been omitted for privacy, or attracted little activity.
Similarly, a URL disappearing from the Pages export is not evidence that it was deindexed. Verify its current indexing state, canonical attribution, redirects, and actual Search Console filters before deciding what happened. Treat the export as a selection of reported rows, not an exhaustive inventory of all your rankings.
Why CSV Totals Might Not Match the Search Console Chart
Google states that direct report exports are limited to 1,000 table rows of representative data, while report totals can include additional activity. Search Console also omits certain rare queries to protect privacy. Query rows and page rows can use different aggregation methods, which means their totals are not always interchangeable. Consequently, adding uploaded CSV rows might yield a lower number than the headline clicks or impressions displayed in Search Console.
For extensive historical or large-site analysis, evaluate Search Console API extraction or BigQuery bulk export instead of assuming a manual CSV is complete. Do not describe a difference between summed CSV rows and Search Console chart totals as a data corruption error without first accounting for reporting limits.
Interpret Declines as Clues, Not Automatic Diagnoses
A decline in clicks combined with lower impressions could be associated with reduced search demand, ranking visibility, query mix, or changes in available Search features. Rising impressions with falling CTR can reflect broader visibility on lower positions, competing results, or a snippet that attracts fewer clicks. Position changes can also reflect changing query mixes. The CSV alone cannot prove an algorithm penalty, an AI Overview effect, or that a title rewrite caused the change.
Sort substantial losses by absolute clicks first, then check their impressions, CTR, and position together. For an important declining URL, inspect its ranking queries, current title and snippet, page updates, canonical status, and nearby competitors. For branded versus generic searches, use a consistent custom segment, potentially prepared with the Search Console Regex Builder and Tester.
Check File Quality Before You Trust Any Comparison
- Column headers: confirm which field identifies the query or page and which columns contain clicks and impressions.
- CSV structure: quoted values, commas inside phrases, and different regional number formats can affect parsing.
- Repeated keys: if there are duplicate query or page identifiers, aggregate their counts deliberately rather than silently keeping one row.
- Percentages: distinguish a stored CTR value such as
3.5% from a fractional value such as 0.035.
- Missing values: do not interpret a blank position or zero-impression row as a real ranking at position zero.
- Privacy: review whether exported search queries contain commercially sensitive information before sharing files with third-party websites.
Processing location, row limits, supported header languages, automatic file downloads, and accepted comparison-export formats depend on the particular analyzer. Check the tool interface for those features instead of assuming every CSV comparison service implements them.
Build an SEO Review From the Largest Verified Movements
Use the comparison to create a short, evidence-based action queue: pages with material click losses, queries with growing impressions and weak CTR, and content groups moving in opposite directions. Check whether the movement spans several queries or one term. A single row with a dramatic percentage change can be statistically trivial if it moved from one click to two.
Finally, document the dates and filters and repeat the analysis after a reasonable observation period. For authoritative details on metric definitions, exports, and known reporting limitations, refer to Google’s Performance report guide, Search Console export documentation, and Google’s data discrepancy guidance. A comparison is most useful when it produces a testable editorial or technical question, not a speculative explanation.
