Colquhoun.AI
← All guides

Research Guides

Report Branded Search Without Hiding the Unknowns

· Colquhoun.shop editorial

A report says “70% of our search traffic is non-brand.” That sounds precise, but what was divided by what? A list of visible search terms is not necessarily a complete account of every click. A brand-matching rule also makes judgments: a company name may be a common word, a product may have several spellings, and some queries may name both the company and a competitor.

This guide provides a reproducible classification record for your own Google Search Console exports. It does not describe a Colquhoun AI integration or claim that Colquhoun can retrieve private search data.

Start with the reporting question

Separate two questions:

The first is a classification exercise. The second needs a compatible total and evidence about missing detail. Do not silently answer the first and label it as the second.

Google documents several reasons why chart totals and table sums differ: anonymized queries, row limits and aggregation differences. Query filters can exclude anonymized queries, and matching plus nonmatching filtered totals need not reproduce the unfiltered total. Therefore, the difference is not automatically “non-brand traffic.” Google’s data-discrepancy guidance.

Write a brand dictionary before calculating percentages

Create a small decision table with four columns: term or pattern, intended meaning, counterexample, and treatment. Use real knowledge of the organization, not guesses derived from the traffic chart.

For a fictional company, Cedar Lantern, a first pass might include:

Candidate query Proposed treatment Reason
cedar lantern login Brand-related Clear company-navigation intent
cedar lantern reviews Brand-related Explicit company reference
cedar lighting ideas Not matched by the dictionary Shared word alone is insufficient
cedar lantern vs northstar Brand-related, comparison subtype Includes the company and a competitor
lantern Ambiguous Generic product word

These examples are an editorial classification scheme, not Google categories. Keep ambiguous rows separate. A reviewer should be able to disagree with one decision without rebuilding the whole report.

Record known spelling variants, old names and product names only when their relationship is established. Do not add a misspelling simply because it produces a large number of clicks.

Test the rule on both matches and nonmatches

Google’s custom query filters support RE2 regular expressions and default to partial, case-insensitive matching. Anchors change where a pattern can match. This matters when a short brand name also appears inside unrelated words. Check the actual matching rows before trusting the total. Google’s filtering documentation.

Review a sample of high-click matches, low-click matches and excluded terms. Keep the original export unchanged; put classifications in a separate column or worksheet. Save the exact rule and its version with the result. If a rule changes next month, rerun both periods with the same rule or disclose that the definition changed.

Use a denominator you can name

Here is a fictional, internally consistent export:

Within the exported rows, brand-related clicks are 200 ÷ 800 = 25%. Other classified clicks are 68.75%, and ambiguous clicks are 6.25%. The visible rows account for 80% of the headline total. The remaining 200 clicks are not explained by this export.

Do not label all 800 clicks outside the brand bucket as non-brand. That would combine classified non-brand, ambiguous and unavailable detail. Nor should you distribute the unexplained clicks proportionally among the known categories: that creates an assumption, not a measurement.

If the headline and row totals use different scopes or aggregation, stop before calculating the 80% comparison. Label the two figures separately until their compatibility is established.

More rows are not a promise of complete query detail

Moving a report to an API is not a reason to remove its limitations note. Google says requests grouped by page or query may lose some data; its guidance distinguishes detailed extraction from requests designed for accurate aggregate counts. Preserve the request dimensions, filters and aggregation mode alongside the output. Google’s Search Console API extraction guidance.

Copy this handoff note

For [property, report, dates and filters], rule [version] classified [B] of [Q] clicks in the available query rows as brand-related. [A] clicks remained ambiguous. The separately recorded headline total was [T], with [compatible/not yet comparable] scope. Unavailable query detail was not reassigned to either brand or non-brand. Next review: [specific ambiguous pattern or extraction check].

Attach the dictionary and export to that note. The useful result is not the cleanest-looking pie chart. It is a classification that another person can reproduce, challenge and improve without inventing the missing searches.

This guide describes the built-in Web Reader and Keyword Density behavior verified for this publication. Available data depends on the page response; these tools do not measure visits or search rankings.