Two research students working with Google, Maryam Kamvar of Columbia University and Shumeet Baluja of Carnegie Mellon University, have written a paper which provides an insight into users’ mobile search behaviour. I’ll talk about some of the parts I found most interesting below, but I’d highly recommend you read the full text of this insightful work here.
Google provided Kamvar and Baluja with a random sample of 1m page queries made by the customers of single US carrier in one month of 2005. The data contained queries generated both by Google’s XHTML interface (designed for mobile phones with 12-key keypads) and its PDA interface (designed for devices with QWERTY or stylus input). This enabled the authors to look at the behavioural differences of these groups when performing search queries in the mobile environment.
It surprised me how little difference there was between the length of search terms entered by both groups of users. I would have expected those with better input capabilities to be more detailed in their queries, but there was actually a surprisingly small increase in the number of words and characters used by PDA owners. They averaged 2.7 words and 17.5 characters per query, just slightly ahead of the 2.3 and 15.5 recorded by 12-key mobile phone users.
In fact, even desktop users average out at only 2.1, leading the researchers to suggest 2 – 3 words may be a ‘ground truth’ of all web search mechanisms, mobile or otherwise.
It appears both groups of users are also very selective when it comes to clicking through to view pages found by the search engine. The authors don’t provide a specific figure, but characterise the number of clicks resulting from search as ‘consistently low’ and suggest this is because most users are simply relying on ‘snippets’ of information found on the first results page.
Analysis of the data also provided a fascinating chart which shows the relationships between different categories of search. Adult content was by far the most popular, representing 20 percent of requests. This compares with less than 10 percent for desktop searches. The chart enables readers to look at a particular category of content and see how many users then went on to make a query in a related category. For instance, 13 percent of users looking for real estate information then went on to search for local services in the same session.
The overriding trend, however, was for repeat searches to be in the same category, underlining the ‘single mission’ approach to mobile service discovery.
It is worth mentioning the context of this research: it only covers the behaviour of users searching outside the operator portal and therefore may be skewed towards more technically-literate customers.
There are some general conclusions I arrived at after reading the paper. Mobile search is very much about mission rather than exploration (for more on this, see my previous article of that title). The fact few users click through from the results page can be attributed to a desire to find a small piece of information, often a phone number or address, in the minimum amount of time – it differs from the immersive search, explore, refine and browse approach seen on the desktop.
Users also seem to have specific services in mind when they search. This is as much a service discovery mechanism as an information retrieval tool. The authors found that some 17% of 12-key queries were actually web URLs and the user had mistaken the search box for a web address bar. I don’t know whether that reflects more negatively on the carriers for forcing such a high proportion of users to go through such a torturous process to escape their portals or the handset interface designers for failing to provide a simple URL mechanism in the browser application?