Most of my readers are marketers, and probably weren’t at WWW2008 on April 25 in order to see the presentation of the brilliant new study of mobile queries, Deciphering Mobile Search Patterns: A Study of Yahoo! Mobile Search Queries by Yi, Maghoul and Pedersen of Yahoo!. I wasn’t either, but since we now have access to a PDF of the report, we can see what we missed. Here’s a high level overview for those of you who, like me, aren’t computer scientists, but are still interested in how mobile users search.


Why This Study is Notable

Why should marketers be looking at this study rather than the studies of mobile queries that came before? Several reasons:


  • At 20 million queries over the course of 2 months, this is the largest study of mobile user search queries ever. Kamvar and Baluja’s 2007 study Large Scale Study of Wireless Search Behavior, by contrast is about 1 million queries over 1 month. A larger sample size means more and better data.
  • Queries across several applications are analyzed, including the Yahoo! Go java app, the XHTML/WAP site, and SMS queries. Using this data, a marketer could do mobile-specific keyword research based on the type of user they are looking to target.
  • Gives insight into Yahoo!’s mobile ranking algorithm
  • Compares US-based English queries to International English queries. It’s possible for marketers to do mobile-specific keyword research based on characteristics of queries for US-based and international English-speaking audiences.


Best Practices for Marketers from this Study


This was written for a more technical audience, but there are a number of takeaways for marketers looking to better target mobile users.


1. Application-Specific Search Behavior – Probably the most interesting thing about this study in terms of its contribution is the revelation that users search for different categories based on the type of search they’re doing. For example, sports and travel searches were more common in the widget-based Yahoo! Go than in the XHTML/WAP based version of Yahoo! OneSearch. Entertainment searches (which in this study include adult queries), accounted for 55% of XHTML/WAP based queries, but only 14% of SMS queries. SMS queries also skewed toward sports and travel, but also had more retail-specific queries than any other application. What this suggests to marketers is that mobile search marketing extends way beyond WAP sites into widgets and text messaging as well. When doing mobile-specific keyword research, marketers would do well to consider which application is most trafficked by their user base and target it first.


The authors of the study are planning a voice recognition study that may uncover even more variations in subjects.


2. Yahoo! Mobile Results Factor in Concept and Intent – If you’re still putting all of your SEO eggs in the keyword density basket, you’re probably going to have no chance at all to rank in Yahoo! Mobile. According to the paper, oneSearch determines relevance not just by analyzing keywords, but by looking at the concept and intent behind those keywords. It’s not detailed in the paper how it’s done, of course; but 1) it suggests that the semantic search promised by Search Monkey to be coming is to some extent here today, and 2) it makes me feel that much better about the new Search Behavior Analysis that we developed at Resolution Media to take keyword research into the 21st century, as it factors in concept and intent among other things. Keyword density, after all, while somewhat useful for competitive research, has been more or less irrelevant in ranking for a while.


3. Entertainment Content is King – This study confirms what previous studies found, which is that searchers are largely looking for subjects that could be categorized as entertainment. Unlike previous studies, this study categorized adult queries as belonging to the entertainment category. If your content can be categorized as entertainment and it’s not accessible to mobile users, this should be a call to arms. If you’re considering building a mobile site and you’re looking for a focus or selecting content for the mobile site, entertainment is a good way to go. Sub-types of queries for entertainment included music, movies, games, TV, “amusement”, adult and romance, radio, performing, etc.


Other popular top level categories included people, retail, sports, technology, and travel categories. The study goes into detail about percentages and subcategories for each.


If you’re still thinking of local content as being a primary driver in mobile search, you might be disappointed by this study, which reported local queries as being only 10% of the total sample data. This is considered as a meta-category by the study authors, which spans several top level categories.


4. More Spelling Errors, Long Tail Require Paid Search Supplement– This seems like a no-brainer given the difficulty of query input, but this is the first study that I’ve seen that reports a “higher rate of spelling error [sic]” due to “the difficulties of mobile device input user interface”. SEO probably isn’t the ideal solution for most brands when it comes to optimizing for spelling errors, but it’s a great opportunity to capture those clicks through paid search. Appearing for spelling errors using mobile search advertising may be more important than in the desktop web given the difficulty of query input.


Furthermore, given the find that there are a few queries that are repeated many times, there exists, as in web search, a “long tail”. Optimizing for every query in that long tail may be impossible in natural search, but relatively easy in paid.


5. Mobile Queries Not Inherently Shorter – Average query length for this study was 2.35 words, which is in line with previous mobile query studies. However, in Google’s 2007 study, the most frequent word count was one word, while for this study the most frequent word count was 2. The authors of the paper speculate that the similarity to desktop web search queries of mobile search queries indicates that “mobile users repeat some of their search habits from desktop search on mobile search”. For marketers, this means that doing mobile keyword research doesn’t necessarily entail using the shortest version of the query to be visible in mobile search.


6. Queries Less Navigational – As I noted in my study of head queries from the Jumptap Mobile keyword tool, the queries were disproportionately navigational in nature. The authors of this study note that navigational queries made up only 5% of the sample data, as compared to 17% of the Google study. Head terms tend to be more navigational than others, so if marketers are looking for volume, navigational queries are still a safe bet.


7. Slight Differences Between US and International English Queries – Though query subjects between US-based and international English queries were largely the same, there were slight differences between queries of the groups in the structure of the query. For instance, US-based queries were 11% longer than their international counterparts. This might seem intuitive that non-native English speakers use shorter queries given their relative lack of the command of the language, but it is one aspect of mobile search that varies by region. If you’re a marketer doing international campaigns in English, it’s helpful to do mobile-specific keyword research for US and international users.