Natural language testing on AdWords expanded text ads

By Dom Ebel


Natural language testing on AdWords expanded text ads

As of 31 Jan 2017, Google stopped the creation and editing of standard "text" ads - enforcing the creation of the expanded text ad format introduced in May 2016. The expanded text ad format uses 2 headings to create a longer conjoined heading, with a single description line, to produce a uniform format working over a range of devices with differing screen sizes.

One of the smaller changes seen with this new ad format is that keywords in the heading matching the search query are no longer bolded, a feature that used to highlight ads by drawing the viewers eye to ad messaging that matched their search query.

While reflecting the users search query wording is still important, with no bolding to draw the eye to ad headlines reflecting the matching keywords we wondered - might this be the time to give ourselves more lateral space to create headings that describe the offering in a more accurate and compelling manner, rather than constraining ourselves by trying to reflect the keywords used in the search query?

While not departing hugely from the keywords employed in the search query, we constructed some tests comparing ads with identical description lines and domain paths but different headings - one set using the highest-performing ad copy variant reflecting the search query, and a second variant crafted to describe the product offering in the most accurate manner possible using a slightly more “natural language” description (with an eye to user search query keywords, but not constrained by them).

We created an ad copy test running across nearly 300 ad groups, for a period of over 100 days, and while the results were not as clear-cut as we would have expected, analysing the results ad groups with statistically significant volumes of traffic showed the following:

  • “natural language” ad messaging had a higher click-through rate 78% of the time
  • the average performance for “natural language” ads was an increase in CTR of 0.86%
  • the highest increase seen (a brand ad) was of over 5% CTR above the best-performing variant that reflected the search query
  • where the “natural language” ad performed best, the average increase was 0.95%

The bottom line – we have seen an average of a near 1% increase in CTR by using natural language messaging in the ads, in some cases higher.

As a result, while not necessarily appropriate for all situations, this is an approach we are now starting to apply as part of our ad copy strategies to help improve client PPC performance.