How to validate a Search Console export before analysis
You export Search Console data, open the spreadsheet and there it is: a query that looks like a big opportunity. You’re already planning the slide.
Hold on a moment. Before you build a report around it, check what the export actually contains. A neat table can still be incomplete, filtered differently from the view you meant to analyse, or mixed with rows from another date range. Spreadsheets don’t warn you about any of that.
Confirm the export context
Write down the property, report, search type, date range and filters used to create the export. Search Console has different property types and report views, and a page-level filter can change which queries appear. If someone else prepared the file, ask for the exact settings. Don’t guess from the column headings. I write the settings into the file name of every export I save, because I don’t trust myself to remember them a month later.
Check the headers, date format and row count, then look for blank values, duplicate rows and totals that do not match the on-screen report. When combining exports, make sure each uses the same dimensions and date window. Do not assume that a missing query means no one searched for it; privacy filtering, reporting limits and selection settings can affect what is shown.
Validate the data before interpreting it
Choose a few sample rows and confirm that the page URL and query make sense for the property. Check whether URLs include parameters, alternate protocols, hostnames or trailing slashes that need normalising. Preserve the original export, and perform cleaning in a separate copy or worksheet so that you can trace every transformation.
Understand the metrics before calculating ratios. Clicks and impressions have straightforward definitions within the report, while average position is an aggregate that can hide differences across queries, devices and result layouts. Click-through rate is useful for prioritisation but needs context: a result feature, brand query or page intent can change expected behaviour. Don’t treat one average as a universal benchmark. Average position, in particular, is the metric I’d least like to see on its own in a client report.
If you calculate totals, use formulas that can be checked and make exclusions visible. Avoid counting the same URL twice after normalisation. Do not combine page-level and query-level rows as if they were equivalent. Keep a note of any rows removed, grouped or renamed.
Connect analysis to a question
Once the export passes basic checks, ask what you are trying to learn. A page with rising impressions and flat clicks might warrant a review of the search result and query mix. A fall in clicks could reflect seasonality, a changed result layout, a technical issue, a content change or data boundaries. The export alone will not identify the cause.
Compare consistent periods and account for the site context: promotions, stock, migrations, page updates and market demand. If you compare before and after a change, record what else happened during that window. Treat patterns as evidence for investigation, not proof that one edit caused a movement.
Keep a reproducible record
Save the untouched export with its date and source, the cleaned analysis separately, and a short note explaining filters and calculations. This makes it possible to repeat the report and spot changes to the process. If the figures feed a client or business decision, have another person review the transformations and definitions.
Before presenting a conclusion, can you explain where every key figure came from and what the export cannot tell you?
Avoid false precision in spreadsheets
Exports are often joined to analytics or a keyword list. Before doing so, define the join key and inspect whether it is unique. A page URL may appear across many query rows; joining it to another one-to-many table can multiply totals. Keep the original data intact and test calculations on a small sample where the expected result is easy to reason about.
When you share a spreadsheet, label calculated fields and explain the formula. Protect against accidental edits to the source tab, and remove any information that should not be distributed. A clear workbook should let someone distinguish raw data, cleaned data and interpretation without needing to reverse-engineer every cell.
Interpret zeroes and missing rows carefully
A zero value, blank cell and absent row do not necessarily mean the same thing. Check how the export represents missing data and whether a filter has excluded the page or query. Keep that distinction in the report. It is better to say “not shown in this export” than to state that a query received no impressions when the data does not support that conclusion.
Large reports can be truncated by a tool or export method. If totals do not reconcile, consider whether the interface limits rows or whether the file was cut off. Compare an aggregate view and a sample of detail rows, then explain any known limit. Do not silently combine partial files and present the result as complete.
Preserve a review trail
For each important finding, retain the formula or filter that produced it and a link to the source file. If a colleague challenges a number, you should be able to reproduce it without relying on memory. A second reviewer can check whether the processing steps match the question and whether the conclusion is appropriately cautious.
This may feel like extra work for a small analysis, but it saves time when the report is revisited months later or handed to another person. You’re not aiming for a perfect spreadsheet. You’re aiming for an honest chain from export to decision. Which assumption in your current Search Console analysis would be hardest for someone else to verify?
Keep context beside each chart
A report should make definitions easy to find. Add a note that explains the property, query filters and date range directly beside the visual, not only in a separate email. If a chart shows an average position, explain that it aggregates varied queries and does not represent a fixed rank every searcher sees. That small bit of context makes the chart much harder to misquote when it gets forwarded.
Related reading
- Search Console exports for product page patterns
- The 24-hour view in Search Console
- Search appearance filters in Search Console
Next time you export, try this before any analysis: write down the property, report, search type, dates and filters, then spot-check five rows against the screen. Does everything still match?