Your Website's Search Interface: WebRef Update Feature | WebReference

Your Website's Search Interface: WebRef Update Feature

Collecting Feedback About Your Website's Search Interface

It's crucial for websites to provide search interfaces that are available, simple, and productive. This article gives basic instructions about how to test your website's search interface for usability.

We recently assessed the usability of twenty ecommerce sites: ten of the world's highest-grossing sites and ten other sites that sell the same kinds of products, but have substantially smaller revenues. The ten high-selling sites complied with 40% of our usability guidelines for search whereas the ten less-selling sites only scored 27%. Even the best sites on the Web are far from having perfect search, but it is still remarkable to note the difference in search quality between sites that sell well and sites that sell poorly. Of course, rule #1 of ecommerce is: if the user cannot find the product, then the user cannot buy the product. Thus, the findings in our study are very understandable. Get more information about the study at:

Search is something most people try to use on websites. Some people simply prefer it to browsing. Others like to browse but are driven to using search when browsing does not help them find what they want. Often the information on websites is organized in categories, and the user does not know which category a particular item is under. The site designer may have a logical structure in mind, but the arrangement still might not help the user. For example, a person wants to buy a Lenny Kravitz CD for his nephew but he does not know what type of music Lenny plays. He might try browsing a music site and looking in the classical or country categories. He wouldn't find the CD there. He would probably resort to searching the site for the artist's name, or possibly a song title or album name if he knew it. If that didn't work, it's likely that the shopper would leave the site and go to another one to make the purchase. To avoid this catastrophe, web designers can conduct basic usability studies to help them learn how to improve their search usability.

Steps in a Basic Usability Study

Usability studies can be elaborate or basic. Even 30 minutes in a cubicle with paper and a pencil can be effective. When looking for quick feedback about a search interface, your study should include the following steps:

1) Determine the goals of the study 2) Determine the user profile and schedule the sessions 3) Write the user's tasks 4) Conduct the sessions 5) Evaluate the data 6) Implement changes, and repeat steps one through seven as long as the site exists

Determine the Goals of the Study

Decide on some clear, concise goals for the study. For example: Learn what is easy or difficult about constructing a search. Or, learn what is easy or difficult about using search results. Keep these goals in mind when writing the user tasks.

Determine the User Profile and Schedule the Sessions

The people who test the interface should absolutely have similar experience levels as those who will use the live site. Decide what traits are pertinent, for example: web experience, search experience, occupation, age, experience using competitive sites or the site you are testing.

Be prepared to offer an appropriate incentive. Some popular honorariums include cash or a check (often $100 US), sweatshirt, software, pens adorned with a product name.

Write the User's Tasks

While there are many steps and features involved in searching, there are a few basic things people should be able to do, and that you should try to include in your user tasks.

1) Find the search interface: Do not bury it. 2) Type words, create search string: Search should accept and understand both natural language and Boolean strings. You will probably find that most users simply type a single keyword or at most two or three words. Complex queries are rare on most sites. 3) Invoke the search: Provide a clearly-labeled button near the query entry field. People expect the Enter key to work too. 4) View, use, and understand the results: The user needs to know where the results are, not buried in categories or advertising, and which results are the most relevant. Also, opening and reading the pages or documents should be straightforward. 5) Refine or narrow the results.

In the tasks you would give the users, you would not actually use the above words. On the contrary, these are just parts of the bigger user task or goal to "get the information I want," or "buy Mom a birthday present."

During usability sessions, we often give users a protocol or set of tasks to follow. Writing the tasks is more of an art than a science. The idea is to give people enough information without guiding them to the solution. Tell them things they should try to accomplish that are realistic and might occur in a real-life scenario. For example, one task might be: Get your mother a birthday present. In this task you don't use the word search, you don't tell them any product names, you don't tell them to deal with results or relevance rankings. You leave it open-ended for the user to decide what Mom would like, and how they will find it. Then, you watch them try to get the work done, and analyze their behavior as they do.

Next Page

This article originally appeared in the October 26, 2000 edition of the WebReference Update Newsletter.

Comments are welcome
Written by Jakob Nielsen and Kara Pernice Coyne and

Revised: Oct 27, 2000