Define what counts as your brand
Start with a written definition rather than an improvised filter. List the company name, distinctive product names and spelling variants that your team can justify. For each entry, record why it belongs and who reviewed it. Do not add every service word simply because the company sells that service.
Decide how to handle names that are also ordinary words, shared abbreviations and queries mentioning both your brand and a competitor. Keep an ambiguous category in the working review when the text alone does not resolve the meaning. The objective is a consistent analytical rule, not a claim that you can read every searcher's intention.
Translate the dictionary into a tested filter
Google supports query substring filters and custom regular expressions. Its documentation says regex matching uses RE2 and is partial by default; anchors can restrict the match. Choose the simplest expression that implements your definition and keep a copy of the exact text used in the report.
Test the rule against a small reviewed list before applying it to a client summary. Include obvious brand queries, spelling variants, ordinary service searches and misleading near-matches. For an Arabic and English business, review observed spellings in each language rather than assuming a translated company name covers the ways customers actually type it.
Example: a distinctive name and a misleading substring
Illustrative fictional brand: LumaNest. A simple rule matching 'lumanest' can identify 'lumanest support', but it misses 'luma nest support'. Broadening the rule to 'luma' may also capture unrelated text such as 'lumatherapy'. Add the spaced spelling deliberately rather than expanding the match beyond the agreed brand vocabulary.
A teaching expression such as (lumanest|luma nest) addresses those two written forms, but still needs review because partial matching can include longer strings containing them. Record expected matches and exclusions next to the expression. This invented brand demonstrates a test method, not a universal regex to copy into every account.
Keep the comparison scope and uncertainty visible
Use the same property, dates, search type, page scope, country and device settings for both segments. Save the filter rule with each export so another reviewer can reproduce it. If the dictionary changes, rerun both comparison periods using the revised rule instead of treating the resulting difference as market growth.
Google warns that query filtering can change totals because of truncation and anonymized queries. Label your result as a classification of the available filtered data. Do not force the segments to equal an unfiltered headline by allocating an unexplained remainder to nonbranded demand.
Turn the split into a specific next question
If branded activity changes, investigate the relevant brand terms and landing pages before assigning a cause. If nonbranded activity grows, examine whether the queries describe services you actually offer and whether their destinations answer the question. Segment labels alone do not establish lead quality or explain why traffic changed.
Maintain a versioned dictionary with its review date, expression and ambiguous cases. Include a few manually checked examples in the report appendix. This makes future updates manageable and prevents a quiet filter change from appearing as an SEO success or failure in a monthly chart.