August 30, 2010 |
SharePoint Search / FAST Search Implementation Pitfalls - Part 1 Enterprise Search best practices are often hotly debated, depending on who you speak with. A System Integrator will often argue that a successful search implementation starts with a complete assessment of an organization's information architecture, whereas an ISV will argue that software should (and is capable) of automating much of the manual labor associated with optimizing an enterprise search platform.
So rather than present my perspective, which I'm sure you can guess (BA Insight is an ISV), I thought that presenting some findings from over 20 years of research on what has proven not to work would be of interest to some of you.
I'll start with a few search "features" I've seen that have been made available by the open source community. At first glance, some of these look wonderful and create a lot of value, but history tells otherwise.
Search As You Type
This feature presents users with a dropdown beneath the search box containing possible hits based upon what has been typed to that point. Clicking on a particular value will bring the user to the content bypassing the search results page.
This feature looks very similar to a feature seen on Google which suggests possible query terms to the user as they begin typing a search. In search parlance the Google feature is referred to as Query Reformulation. This refers to the fact that users must often change their initial query because it is often difficult to provide the right keywords on the first attempt. Google is presenting the user with alternatives as they begin to type a query potentially saving the user the trouble of having to reformulate.
Examinations of search engine logs suggest a high frequency of query reformulation. One study by Jansen et al. (2005) analyzed 3 million records from a 24 hour snapshot of Web logs taken from the Alta Vista search engine. The analysis found that users who modified queries were 52%, with 32% issuing 5 or more queries within a session. Other studies show similar numbers, indicating query reformulation is a common part of the search process. So this sounds like a great feature to implement...What's the problem?
The problem is the Search As You Type feature illustrated above IS NOT Query Reformulation. What Google presents is a list of alternate queries. What the web part above presents is a search result. The user must know exactly what they are looking for based on the limited information being presented in the dropdown. This bypasses the search results page which is designed to help the user in countless ways (Faceted Search, Document Previews, Document Summaries, etc.). This is a huge step backwards in terms of user experience, and disregards much of what has been proven effective in search over the years.
I'm sure someone reading this will still say that there are certainly situations where this could still be useful, but ask yourself...How often do you click the "I'm Feeling Lucky" button on Google. In the next post I'll look at several approaches at Query Reformulation. Some work very well. Others provide questionable value.
|
|
|