If a brand runs a television campaign, branded searches rise, total clicks rise, and the SEO report shows growth that SEO did not cause. Splitting brand from non-brand is what makes the rest of the report trustworthy.
Build the brand list carefully
Start with the obvious: the brand name, its common misspellings, and the domain with and without the TLD. Then add the things people actually type — the brand plus "login", the brand plus "customer care", product names that only your company uses.
Be careful with generic-sounding brand terms. If the brand is a common word, a naive match will pull in genuine non-brand demand and flatter the wrong column.
Match on the query, not the page
Branded traffic lands everywhere, including on pages you consider purely informational. Classifying by landing page will misattribute a large share of it. The query is what tells you intent.
The totals will not reconcile
Search Console withholds queries that are too rare to report. Those clicks appear in your property total but in neither the brand nor the non-brand bucket, so the two will always sum to less than the whole.
That gap is normal at a few percent. If it grows past ten, either your row limit is truncating the tail or the property has an unusually long tail of one-off queries. Report the gap rather than quietly distributing it.
What to do with the split once you have it
Non-brand clicks are the number that reflects SEO work. Track it as the headline and keep total clicks as context. When non-brand rises while brand falls, that is a good month for search even if the total looks flat — and being able to say so, with the numbers behind it, is the whole point of doing the split.