Search Query data is very useful information – it’s what the user actually typed into a search engine.
It’s often used in two ways:
While those are the two most common reasons to use them, the problem with that analysis is that when search queries don’t have many clicks, they are just ignored by advertisers.
For someone more sophisticated, they manage low volume queries by examining interaction rates (such as bounce rate or time on site). For instance, if you see a word has 20 clicks, it will generally get ignored. However, if you saw it has 20 clicks and a 100% bounce rate, then you’ll give it less leeway before making it a negative.
The disadvantage to just using interaction rates is that it doesn’t look at patterns across your data – it still just examines the actual query.
What advanced advertisers do is use n-grams across query data to find insights in their data.
N-grams analysis examines the data at the individual word or term level and aggregates the data across all search queries using that pattern.
For instance, in this account the word ‘napkins’ occurs in 1804 different queries and when that word appears in the query they have a 4% conversion rate; so napkins is an important word for them.
Now if we keep digging through high spend and low conversion words (just filtering by words with less than 2 conversions and sorting cost high to low) we see that the word ‘wine’ has appeared in 99 different queries and that when it appears there is only one conversion at a very high CPA.
You might ask yourself why this word wasn’t spotted before and added as a negative. Well, when you look at the actual queries, none have more than 60 clicks and only one even has 20 clicks. This is the weakness of just looking at query data – you don’t spot patterns than when N appears in your search query that the advertiser isn’t doing well.
Now at this point in time, we can discover:
Next, we have two options:
Often we examine search queries for finding negatives or new keywords. This is a necessary part of managing your account. However, examining search queries one by one doesn’t lend itself to data insights.
In this case, you would have ended up with at least 99 negative queries (there are 99 distinct search queries with the term wine) to block all of this traffic instead of just making a single negative keyword (-wine) to block everything since you can see the root cause of this poor traffic.
In addition, using n-grams allows you to not only find new negatives, but think about how you can expand your businesses to meet search demand.
If you want to learn more about how n-grams work, please see this article.
If you want to see n-grams for your own data, take a free trial of Adalysis.