Branded vs Non-Branded Search: Measure Demand Without Hiding the Gaps

Define branded, non-branded and ambiguous queries

Write a short brand dictionary before extracting a report. Include your trading name, domain, common spacing variations, distinctive product names and verified misspellings. Confirm that each entry actually identifies your business in the markets you serve.

For a fictional company called Acme Metrics, “acme metrics pricing” clearly contains the brand. “marketing reporting software” is non-branded under that dictionary. “acme” can name several businesses, so classify it as ambiguous until the query and landing page give you enough context.

A query containing your competitor’s name is not automatically your branded demand. Keep competitor comparisons as a separate label inside your analysis if they matter to your content plan. Document how queries containing both your brand and a competitor are handled.

Keep the dictionary version with the report. If you add a new product name halfway through a monthly comparison, rerun both periods under the revised rules before describing the change as growth.

Choose Google’s native classification or a reproducible custom rule

Google’s branded-query filter uses an AI-assisted classification system that can recognise names, typos and distinctive products. It is different from including or excluding a list of strings. Google also documents eligibility limits and possible classification errors.

Use the native filter in Search Console when it is available and fits your reporting question. Inspect examples before treating its categories as final. A native branded total and a custom regex total can legitimately differ because their classification methods differ.

The current SourcetoMCP performance operation exposes text and regex query filters. It does not expose a native branded-classification parameter. Ask your assistant for a custom-rule analysis, or supply an exported native report if that is the evidence you want it to interpret.

For a custom analysis, request query rows and classify each returned query using your reviewed dictionary. Inspect uncertain matches manually. This produces an auditable classification of the returned rows, with missing and anonymised queries still outside the classified sample.

Apply query filters and test what they include

The connector supports includingRegex and excludingRegex on the query dimension. A simple starting pattern for the fictional company is acme metrics|acmemetrics. Test the pattern against real returned queries and inspect misses before expanding it.

Broad fragments can match unrelated searches. Whole-query anchors can miss queries that add “pricing” or “login”. Write down whether a rule matches a phrase anywhere in the query or requires the entire query, and check multilingual names separately.

Search Console’s filtering guidance explains regex matching and the effects of query filters. The connector accepts at most five filters, combines them with AND, and caps each response at 1,000 rows. Put alternative brand spellings into one reviewed pattern when they belong to the same rule.

Use the same property, search type, country, device and periods for each comparison. Preserve the filter expressions beside the figures. If a broad query sample hits the row limit, narrow the scope and describe the resulting coverage.

Calculate the share with an explicit denominator

The following hypothetical result contains 1,000 clicks across returned, non-overlapping query rows. The branded and non-branded labels come from a reviewed custom dictionary. Ambiguous rows remain visible.

If a comparable unfiltered, property-level report records 1,250 clicks, the returned query rows account for 80% of that total in this example. The remaining 250 clicks cannot be assigned to either group from this sample. First verify that aggregation and filters are genuinely comparable.

Google documents that anonymised queries disappear when query filters are applied. Row truncation can also affect coverage. Subtracting a branded filtered total from an unfiltered total would incorrectly treat those unknown clicks as confidently non-branded.

Calculate group CTR from clicks and impressions at a consistent aggregation. Do not average row CTRs or positions to invent a property total. Keep sample-based shares labelled as shares of returned query clicks wherever the chart or summary appears.

Interpret growth with the query mix visible

A rising branded share can mean branded clicks grew, non-branded clicks fell, or both changed. Compare the absolute clicks in each group before celebrating a percentage. Inspect the queries and landing pages behind the movement.

For example, branded clicks rising from 300 to 360 is 20% growth. Non-branded clicks falling from 650 to 520 is a 20% decline. The larger branded share in that situation conceals a loss in broader discovery if you show only the percentage.

Compare branded and non-branded Google Search performance for [property] over [two complete periods]. Make no changes. Use this reviewed brand dictionary: [names, variants, distinctive products]. Label the method as custom classification, not Google’s native branded filter. Retain ambiguous rows. Show clicks and impressions by group, the explicit share denominator, exact filters and response limits. Explain absolute changes before percentages. Keep unreported clicks unassigned. Return the queries and pages behind each change, plus specific content investigations. Do not infer query-level sales or attribute brand growth to a campaign without evidence.

Use your SEO report template to preserve definitions alongside the chart. Keep the query groups stable over time, and record any reclassification that changes an earlier result.

Turn the findings into specific content investigations

For declining non-branded queries, inspect whether the destination still answers the question and whether search demand changed. The content-decay guide provides a page-level investigation before you commission a rewrite.

For branded queries, check whether pricing, login, comparison and setup searches reach useful pages. A query for your support instructions should have a clear answer, even if a broad discovery keyword has more search volume.

Search Console query clicks do not identify individual people or provide query-level GA4 conversions. Use landing-page outcomes as separate evidence with their own attribution limits. Keep the proposed content change tied to the question and destination you actually observed.