usdigitalmarketingindex.com SEO
How should branded and unbranded search clicks be reported separately?
Updated November 15, 2026
With a written definition of what counts as a branded query, applied as the same filter every reporting period, and with the queries that cannot be classified shown as their own line. Branded queries are those containing the company’s name, its product names, or recognizable variants of them. People who type them already know the business. Unbranded queries describe a need without naming anyone. The two respond to different things: branded clicks follow awareness, advertising, and reputation, and unbranded clicks follow how well the site competes for topics. Reported together, a television campaign and an SEO project look like one result. The separation is only as good as the rule behind it, so the rule is documented, its ambiguities are named, and it is changed rarely and openly.
Brand term variants
Build the list of branded terms from how people really type the name.
| Type | Examples for a hypothetical company, Northgate Heating |
|---|---|
| Exact name | northgate heating |
| Short name | northgate |
| Run together or split | north gate, northgateheating |
| Misspellings | northgat, norhgate, northgate heeting |
| Domain | northgateheating.com, northgate heating co uk |
| Product and service brands | t200, heatguard plan |
| Names of well-known staff or founders | Where people search for them in connection with the business |
| Former names | The name before a rebrand |
| Abbreviations | nh heating, ngh |
Find variants in the data: export queries from Search Console, sort by clicks, and read the list for anything that refers to the company. Misspellings show up there.
An example query classification dictionary:
| Pattern | Class | Note |
|---|---|---|
| northgate heating, north gate heating, northgateheating | Branded | Core name |
| northgat, norhgate | Branded | Misspellings observed in data |
| northgate | Ambiguous: see below | Also a place name |
| t200 thermostat, northgate t200 | Branded | Own product |
| t200 | Ambiguous | Model number shared with other products |
| heatguard | Branded | Own plan name |
| northgate heating reviews, northgate heating phone number | Branded | Navigational and reputational |
| boiler service northgate | Ambiguous | Could mean the place |
| boiler service, boiler repair near me | Unbranded | |
| competitor names | Unbranded, flagged | Report separately if of interest |
Ambiguous product names
Three kinds of ambiguity need a decision, written down.
Name collisions. A brand name that is also a common word, a place, or another company’s name pulls in queries that have nothing to do with the business. In this example “Northgate” is also a place name.
| Approach | Effect |
|---|---|
| Count every query containing the token as branded | Overstates branded |
| Count only the token with a qualifying word | Understates branded |
| Create a third class | Honest; needs explaining once |
Generic brand names. A company called Boiler Repair Direct has a name made of its main keywords. Queries for the name and queries for the service are hard to tell apart. Record that the split is unreliable for this business, and use the landing page as a second signal: an entry on the home page is more consistent with a search for the company than an entry on a service page.
Brand plus need. “northgate boiler service cost” names the brand and a topic. It is branded, because the searcher chose the company. Some teams split it out as a third class, branded with intent, since it responds to both awareness and content. Decide once.
Exclusion notes to record with the dictionary:
- which tokens are treated as ambiguous and why;
- how brand-plus-topic queries are classed;
- whether competitor-name queries are tracked;
- whether partner or reseller names count;
- the date of each change to the list.
Query filter limitations
In Search Console, the split is made with a query filter. A regular expression lets one filter cover every variant. Its help describes regex search as a way to “filter for, or exclude, multiple queries or URLs”, with values separated by a vertical bar (Search Console help on advanced filtering and comparison).
A filter in that form:
Branded: Custom (regex), matches:
northgate heat|north gate heat|northgateheat|northgat|norhgate|heatguard|t200 thermostat
Unbranded: Custom (regex), does not match: the same expression
Limits to state in every report that uses the split. The first two come from the Search Console help on data discrepancies.
| Limit | Consequence |
|---|---|
| Anonymized queries | Search Console's help says anonymized queries are omitted from the query table, and that when filtering by query "the 'match' and 'does not match' totals may not add up to the unfiltered total". Branded plus unbranded will be less than total clicks |
| Row limits | The table shows a maximum of 1,000 rows; rare rows beyond that are left out of the table |
| Pattern matching is literal | A regex cannot know intent; it matches characters |
| Case and spacing | Write patterns to cover spacing variants |
| Languages and scripts | Brand names transliterated in other scripts need their own patterns |
| Ambiguous tokens | Whatever rule is chosen will misclassify some queries |
Because of the first limit, report three figures, not two.
| Line | Source |
|---|---|
| Branded clicks | Filter matches |
| Unbranded clicks | Filter does not match |
| Unclassified clicks | Total clicks minus the two above |
The unclassified line is real traffic whose queries are not shown. Do not distribute it between the other two by assumption. Its size is itself worth watching: a growing unclassified share means the split describes less of the total.
Where an analytics tool has no query text for organic visits, it cannot make this split from the query. A landing page proxy is possible there, treating home page entries as branded, and it should be labeled as a proxy.
Long-term report consistency
The value of the split is in the trend, and a trend requires an unchanged rule.
| Practice | Reason |
|---|---|
| Keep the regex in a shared document with a version number | Everyone applies the same filter |
| Log every change, with date and reason | Breaks in the series can be explained |
| When the rule changes, recompute history with the new rule | The trend stays comparable |
| Add new product names when they launch | New brand terms would otherwise count as unbranded |
| Review the query export quarterly for new variants | Misspellings and nicknames appear over time |
| Keep the ambiguous class stable | Moving terms between classes shifts both lines |
| Annotate brand campaigns, press coverage, and rebrands | Branded movement has causes outside SEO |
A report layout:
| Period | Total clicks | Branded | Unbranded | Unclassified | Branded share of classified | Rule version |
|---|---|---|---|---|---|---|
| [month] | v3 | |||||
| [month] | v3 |
Reading the two lines:
- Unbranded clicks are the closer measure of what content and technical work achieved, since those searchers had no prior intention of visiting.
- Branded clicks reflect demand for the company. They respond to advertising, word of mouth, news, and seasonality in the business itself.
- A rise in total clicks made up of branded growth is not evidence about SEO work, and a fall in branded clicks is not evidence against it.
- Position and CTR can differ sharply between the two, so averages over both say little. Report those metrics per class.
What the split does not measure is brand awareness. Branded search clicks count people who searched and clicked. They do not count how many people know the name. Choosing measures for an awareness goal is a separate matter, covered in the guide to which SEO metrics are appropriate for a campaign primarily intended to increase brand awareness.
The rule: one documented pattern, applied identically every period, three lines in the report, and every change to the pattern dated.
Add your comment to this answer