All advertisers have a clear and measurable goal for their SEM program: drive more revenue at the lowest possible cost. While revenue generation will always be the primary goal for SEM, data science and text analytics are now able to find a new source of value from SEM efforts that had previously been unobtainable. Lexical Insights is the process of analyzing unstructured text, extracting relevant information, and transforming it into useful business intelligence. In the case of SEM, the hidden value is identifying and quantifying the value of specific words and phrases that drive both positive and negative ad performance (i.e. Word Analysis and Phrase Analysis). Further, Sentiment Analysis can be applied to the same data to find the right tone and attributes that help or hinder performance. To be clear, the data has always been there in the ad performance metrics. However, the analytic framework and results from years of practical application have now refined Lexical Insights so that it has become a valuable and actionable component for many advertisers.

The Results
Below is a typical report for single word CTAs that shows the CTAs that help and those that hinder an ad’s performance. The “impact” of a particular CTA is derived by the absolute value change in CTR weighted by the frequency of that CTA in that ad account. Similar analyses can be run for other word types and for 2, 3, 4, and 5-word phrases. By analyzing the impact of individual words and phrases, a marketer can gain a much better understanding of the language dynamics that are driving an ad’s performance.

Word Analysis - Positive and Negative CTA Words

Sentiment Analysis looks across the same words and phrases used in the Word Analysis above and assesses the tone and type of language that is affecting ad performance. Our analysis of very large volumes of ads across many advertisers has shown that the Sentiment attributes shown below are both the most important and most common Sentiments that drive ad performance. In the chart below, the order of Sentiments indicates the relative importance of that Sentiment for a given advertiser’s account. Further, the words most frequently used within each category of Sentiment are shown to the right.

Sentiment Analysis - Negative and Positive

Actionable Information

A key use of the information from Lexical Insights is to create better ads that improve SEM performance. Being able to identify specific words and sentiments and to quantify their impact is a powerful tool in creating better performing ads. In fact, many advertisers often have a strong desire to immediately run a “find and replace strategy” substituting every occurrence of a low-performing word/phrase with a higher performing one. While using Lexical Insights to improve SEM ads is a primary goal, it is important to do so in a thoughtful manner as we have found that a better approach is to treat each ad group and campaign as unique and to develop some competing hypotheses and test them in various ad groups before making account-wide changes. As often happens in marketing, analytic information is very useful in informing and providing directional guidance that might otherwise be overlooked but human interpretation and common sense are still important. Using this approach and the insight from the Lexical Analysis, we have worked with many advertisers that have been able to significantly and consistently improve their SEM performance.

The value from Lexical Insights extends beyond just improving SEM ad performance as advertisers can use the Lexical Insights information to drive other business strategy actions from product marketing to operations. Many advertisers use the results as part of their internal meetings and reports to inform other areas of the marketing organization of what is working and what is not. For example, a common use of the Lexical Insights information has been to focus and streamline market research efforts to produce results more quickly and at lower cost.

Overall, the data from SEM efforts can be used to provide valuable insight into the impact of the language used to communicate with a substantial portion of an advertiser’s potential customers. These insights enable marketers to develop better messaging strategies that resonate with their customers across all digital marketing channels.

Some lessons that we have learned
We have refined our methodology for Lexical Insights over the last several years and while there are many learnings, some of the key ones are:

  • the analysis requires a lot of data and should be used for larger advertisers with at least 60-90 days worth of data. And more data is even better.
  • the results are unique to a given advertiser and not applicable to others.
  • to maximize effectiveness, the tokenization protocol (for identifying and assigning value to word/grammar types such as CTA, adverb, adjective, noun, punctuation, numbers) needs to be specifically designed for SEM ads.

About AdLabs Technology
AdLabs uses AI/Machine-Learning for ad creation at scale, optimization, and analysis. AdLabs creates hundreds of thousands of ads every year and has optimized billions of dollars in ad spend.


→ Email: