Here is the first in a series of my notes on the 2010 Enterprise Search Summit. The session, Search for Customer Satisfaction at Standard & Poor’s, was lead by Peter Bozzelli, Lead Search Architect, Standard & Poor's and Joe Hilger, Practice Manager, Washington, DC Office, Avalon Consulting, LLC. Here is the session overview. My notes follow.
Standard & Poor’s faces the same challenges as many of its customers, such as the need to search across multiple information repositories and improving relevancy for both highly targeted structured financial data and unstructured, webbased information. From this case study, attendees will find out how S&P identified the key search features that its customers require—from a demand for pinpoint precision to a desire to browse—and learn more about how the team implemented features such as search auto-complete, guided navigation, search applications, and searching different repositories.
Pete led off with some background on S&P. It has been around for over 150 years and has 10,000 employees. It offers a range of financial services beyond its index. Avalon Consulting was founded in 2003 and offers services in enterprise search. In 2008 S&P wanted to improve the usability of their Web site. They did a study and found that search was the top priority for improvement.
The session focused on three of the eight search objectives: easy to use, increased relevancy, and provide a smarter search. The last objective refers to personalizing the search for different user needs. This latter goal was the main focus of the session.
They found two different types of users: ones with specific goals and the ones who wanted to browse and explore. They also had two main types of content: micro-content such as ratings, and longer content such as articles.
To deal with the two types of users and the two types of content, S&P redesigned the Web site with three different search boxes. Each search box was optimized for different goals. There are dedicated searches for ratings (65% of requests) and index information (20% of requests). There is also a global search box that provides a wider variety of content.
In the two targeted search boxes (rating and index) they faced a challenge for relevancy. Their typical user will enter a company and expect to see the company at the top of the list of returns. But there are confounding issues as names can be used for different contexts. They implemented auto-complete to help focus the user on their needs. They also provided ‘exact’ and “related” results. In addition, they included promotions and synonyms, as well as business category sorting. For example, the real name for Greece, a popular search term today, is the Hellenic Republic so they provided both, especially when Greece or Greek was first typed into the box.
Next, Peter discussed the features for the browsing user. They offer dynamically created summaries with search terms highlighted. The summary is based on an interaction of the user’s input in the search box and the content to be more useful to the user. Guided navigation was implemented to filter the user’s actions. They provide suggestions to filter and refine the search. They also aided conceptual search by offering related results even if the exact terms were not included.
S&P increased the feedback on “no results found” pages. They included hints on how to improve search and navigation links back to relevant portions of the Web site. They asked if you meant (term) questions and did auto correct of spelling with a custom spelling dictionary tailored to their content. They wanted to reduce dead ends
In the future, they plan to offer a query tag could of suggestions. If you click on a term, the search is rerun with this term. You can use different metrics to weight the tags. They are also considering social search based on user patterns.