Interpreting Hitwise Statistics on Longer Queries

There was a great session at SMX West recently, entitled 'Latest Stats About The Search Engines'. One interesting statistic that panelist Bill Tancer from Hitwise mentioned was that the average query length to the search engines has been rising. Queries of 5 words or more have increased at a year over year rate of 10%, while single word queries dropped 3%.

Interestingly, there has also been a corresponding drop in 'search engine success', the statistic that measures how successful search engines are at delivering relevant results for the user query. Tancer said that he expects to see both of these phenomena continuing.

Unfortunately, longer queries don't necessarily make the job of the search engine easier. Often longer queries over-specify the search intent, and the search engines are unable to return the relevant results. An example over-specified query is 'Will my rowdy Miniature Schnauzer Spike like vacationing at Dollywood', which, of course doesn't return relevant results, whereas the query 'Are dogs allowed at Dollywood' will give Spike's owner some relevant results to study. 
 
At Ask, we have not noticed a dramatic increase in query length, because Ask has always had a long average query length. (And, why shouldn't we have the longest query length when our search engine's name is Ask?) For a decade, users have relied on Ask to answer their fully specified questions, allowing them to interface with the content on the Internet in a natural way. The users that I meet in my neighborhood and in airports and hotels always tell me that "Ask seems to understand me better." I love hearing that, because it validates a lot of work that we've done in the Semantic Search area.
 
So, what can explain the increase in query length to the other search engines, and also in some of the changes we do see in user queries on Ask?

In a nutshell, users are now expecting search engines to not just index the Internet, they are expecting search engines to process the data on the Internet. Searchers don't consider their query to be just keywords; they are starting to expect that the search engine will understand the intent of the query better. Expressing a query with intent requires more words, and the user's investment of more words means that his or her expectations on the search engine are higher. We are clearly experiencing a transition in the way that people are using the Internet.

This transition is best illustrated with an example. Historically, if a user wanted to find when their favorite actor was going to be on TV, he or she would enter the query 'TV listings' in the search box, then use the search results to navigate to a site containing TV listings. He or she would then browse the TV schedule on the site looking for programs that the searcher recognizes are associated with the actor. This task was time consuming, and inaccurate.

Today, we see queries that illustrate that the user expects the search engine to do all of this work for them. We see queries like 'Steve Carell on TV this week'. Steve Carell is both a TV actor as well as a big movie star, and we see users entering queries that further specify the desired information. For example 'Steve Carell Movies on TV this weekend'. As you can see from these examples, users are now adding temporal qualifiers to their queries. (Try 'College basketball games this weekend' on your favorite search engine, and see what answers you get.)
 
A high percentage of the queries that Ask sees are entered in the form of a question. Sometimes the user is expecting an exact answer to the question. For example, we see students ask questions like 'Can you name the two Senators from Tennessee'. We see questions that don't have a concise answer, such as questions about jobs like: 'How do I become a Washington Redskins Cheerleader' or 'Where can I find accounting jobs in Orlando'.

Web business should take notice of Question/Answering sites that have been built and SEO'd to fill the search engine rankings for these types of user questions (e.g. Q&A aggregators like WikiAnswers, AnswerBag, and Yahoo Answers). While this content is generally very relevant, content directly from companies could be more authoritative. Web businesses may benefit by creating FAQ content that is targeted at answering real user questions about their products.
 
Ask has combed the Web for these types of Question/Answer pairs precisely to find this type of authoritative information for our users. Ask currently has over 100 million qualified Q&A pairs, and we've built a Q&A Search to serve them (Q&A results can also be incorporated into the web results page). Enter the query: 'How does the Amazon Kindle work' in Ask, and look for the quotation marks next to search results to see where the Q&A search content is included.
 
Longer queries present an opportunity for the search engines, because more information is available to understand the user intent, including temporal directives. But, it also means that users will sometimes over-specify the query, and it will be more difficult for search engines to provide relevant results.

This is a challenge that we're glad to continue addressing. SEOs and Web businesses can get ideas for longer question query content by querying the Ask Q&A search with product names, company names, and industry terms. These ideas can lead to FAQ content that may end up in the Ask Q&A channel and in ranking in all search engines.

Thanks for your interest in Ask!

- Keith Hogan, VP, Technology