Virtual assistants, voice commands and artificial intelligence are all driving a huge change in the online search industry. Based on available statistics, 20 percent to 25 percent of searches on Google apps and Android phones are voice-based and Comscore has predicted that by 2020 that number will increase to 50 percent.
The search industry has been trying to dissect and make sense of this statistic in numerous ways. However, Google does not make this evaluation as easy as search queries in AdWords. Google does not show you what is text based and what is voice based. Sure, we can segment the data by device and get a ballpark figure, but then virtual assistants and voice commands are available for use even on desktop. So, in that scenario, what does this exponential growth mean in terms of how we analyze, optimize and expand our online marketing efforts? Here are some thoughts based on a number-crunching effort that analyzed data across AdWords search queries for multiple brands, across retail, real estate, travel and B2B industries.
Queries Are Getting Longer
Queries are getting colloquial and therefore, verbose. What’s interesting though, is that these longer queries are eating into the share of 1-token and 2-token queries. So, over a 12-month period, while the percentage of clicks on 3 and above token queries has been increasing, the same has been on a decline for 1-2 token queries.
Figure 1: Percentage of total clicks based on number of tokens in queries
5Ws and 1H Are Even More Important Now
Virtual assistants have been built in a way that makes you humanize them, so you talk to them like you would to a personal assistant or a friend. Which means if you were looking for the best restaurants for Italian food, your query in Google while typing would possibly just be “best restaurants for Italian food.” However, if you were asking a friend (or your virtual assistant), your query would be a complete question like, “Where can I find the best restaurants for Italian food?”
This leads us to our second learning that the 5Ws (“Who,” “what,” when,” “why,” “where”) and 1H (“how”) are even more valuable now. Along with these, “show me”-related queries (eg. “Show me green shirts”) have been on the rise, too. And, of course, “Okay Google”-related queries, as well. In most cases, over the past year, the queries with these token words have doubled. The maximum increase has been in “where”-related queries, leading to the increased relevance of Google Maps and location settings.
Figure 2: YoY Comparison of queries in a sample account
CTRs Are Increasing Thanks to Longer Queries
Longer queries are usually very specific. In the consumer decision journey, they typically fall in the “Active Evaluation” or “Purchase Decision” stage. For example, “What movies are playing at a theater near me,” “When is the store open,” “How do I get to the store from my location,” or “Where can I find the best pizzas in town.” Not surprisingly, the CTRs for these queries are very high. Based on the analysis, queries with seven or more tokens had the highest CTR compared to other queries.
Figure 3: CTR Changes in March ’17 based on query tokens
So now we are back to the basic question: “What should our next steps be based on these findings?”
- Do not ignore longer queries simply based on low impression or click volume. The total volume of such queries is increasing and there will be a pattern that you could use to add value to your keywords.
- Experiment with questions in ads. See if words like “why, where, how, when and what” can fit in. Analyze if it helps increase CTRs even further.
- Pay close attention to images on shopping campaigns. This is to ensure good consumer experience for “show me”-related queries.
- Set up the right location extensions. The “where” and “how to reach”-related queries would map to your local search ads, including your location extensions.
- Make sure your site is mobile friendly. If you’ve put optimizing your mobile site on the backburner for long, it’s high time you start giving it more than step-motherly attention.
Finally, filter for “Okay, Google”-related queries. Have you made any additional discoveries about targeting voice queries?