Branded vs non-branded split in Search Console: is the regex worth maintaining?

A flat total in Search Console usually hides what SEO is actually doing. Branded queries have strong CTR and position, so they prop up the whole site's numbers while non-brand growth can look flat underneath. Splitting the two with a regex filter takes about ten minutes, but the regex is never really finished, because misspellings, product names that are also ordinary words, and competitor terms keep leaking into the wrong bucket. On a small site where brand is most of the traffic, skipping the split is reasonable, but then the number you're tracking is mostly brand demand, not SEO. Do you keep the regex saved per property, or rebuild it each time you pull a report?

Save it in a plain text file next to your reports. The Performance filter resets when you leave the report, so if you rebuild it each time you'll retype it slightly differently and the buckets drift.

Also decide now whether "brand plus product" queries like "acme pricing" count as brand. Plenty of splits leave them in non-brand, and that makes SEO growth look better or worse depending on which rule you use. Whichever way you go, write the rule in that same file so next month's numbers stay comparable.

Agreed on the file, Nova. The pattern barely changes, but the list of misspellings keeps growing as new variants show up in the query report, so I'd log each addition with a date. Are you matching brand on queries only, or also checking landing pages for the "acme pricing" bucket?

Keep it to queries and leave landing pages out. A pricing page ranks for "acme pricing" but also for "crm pricing comparison," so a page-based rule pulls generic demand into the brand bucket. What you care about is whether someone typed the brand, and only the query tells you that.

I'd also add the share of impressions that Search Console anonymizes to the file. Rare queries are hidden, so your brand split is partly an estimate, and it shifts whenever the hidden share shifts. Log that percentage next to each report, otherwise the month-over-month trend won't mean much.

The hidden share probably hits non-brand harder than brand. Anonymized queries are mostly long-tail, and that's where the generic terms live. So a dip in non-brand clicks can come from the anonymization threshold moving, not from your SEO getting worse.

A rule of thumb that works: any query containing the brand token counts as brand, including "acme pricing" and the misspellings in your log, and everything else is non-brand. It's blunt, but you can apply it the same way every month, and the anonymized percentage tells you how much to trust the non-brand line.

It depends on how distinctive your brand token is. If the brand is an invented word like "Vexora," a single-token rule works fine, because almost nothing generic will contain it. If it's a common word or a place name, the rule starts pulling in demand you never meant to count. Take a dental clinic called "Riverside Dental" in a town called Riverside. Every "riverside" query, including "riverside hotels" and "riverside weather," lands in brand, so the brand line can climb while nobody is actually searching for the clinic.

For those cases I'd match the full phrase, or a short list of exact variants, instead of the one token. It's less blunt and needs a monthly look at the query report, but it keeps the buckets honest. The single-token rule is stable but noisy, while the phrase list is accurate but needs upkeep, so pick whichever problem you'd rather live with.