N-gram analysis
N-gram analysis splits search terms into single words and word pairs and adds up their spend, revealing waste spread across many small searches.
In the platformSearch
N-gram analysis
N-gram analysis splits search terms into single words and word pairs and adds up their spend, revealing waste spread across many small searches.
Words are 1-grams, word pairs are 2-grams, and so on. Waste is often spread across hundreds of searches that each spent a few dollars, so none looks worth acting on alone. Grouped by a shared word such as free, jobs or used, the pattern and its total cost become clear, and one negative keyword can cover them all.
Where you'll see it
Google Ads has no built-in n-gram report. Export Insights and reports > Search terms (and the Performance Max search terms report) to a spreadsheet, or run a script that splits terms into words. Meta and TikTok don't show search terms for most placements, so this is mainly a Google Search, Shopping and Performance Max practice.
Worked exampleThe word used, one month
A furniture store finds 140 different searches containing used, none above $15 on its own.
- Total spend: $640
- Conversions: 1
- Cost per conversion: $640, against an account CPA of $55
One negative keyword, used, as broad match at the campaign level would have excluded all 140. That is an estimate of avoidable spend if the pattern continues, not a measured saving.
Do
- Look at two-word fragments too; how to often carries the intent
- Check conversions, not just spend
- Use a long enough date range
Don't
- Exclude a word that is costly in some searches but converts well in others
- Read small fragments over a short range; they look random