Search Is Not Just an IT Problem
Search Is Not Just an IT Problem
Join the DZone community and get the full member experience.Join For Free
Java-based (JDBC) data connectivity to SaaS, NoSQL, and Big Data. Download Now.
I recently attended the Fall 2010 Enterprise Search Summit anticipating that it would be a battle of the search engines – why choose Solr over FAST, Endeca, Autonomy, Google Search Appliance, etc. However, in talking to people who were there and wrestling with enterprise search problems, I realized that many of them were trying to determine a search engine technology strategy without incorporating that decision into a larger overall search strategy.
A computer will always do what you tell it to do, but not always what you intend for it to do. This is the case in developing a search strategy and for IT teams who are trying to execute the enterprise search technology implementation.
To me, search is about the user. It is a way for a user to find the information he seeks or that he didn’t know about in the first place with the most accurate results in the most rapid manner. Accuracy and relevancy is much more important than speed, as the average user won’t notice the difference in milliseconds in the search results, but will most certainly notice results that don’t further the inquiry.
Thus, the starting point and the keystone of implementing a search strategy is in defining users, both in who they are and what they seek, and also in what action you want from that user. Is your website an e-commerce website that generates revenues from sales, is it a website that generates revenues from advertising and needs long site visits (preferably with many page visits during the user’s stay and subsequent returns), or is it an internal knowledge site where the user is looking to find previous use cases to prevent duplication of work? Does the user have a specific need in mind, or are is he looking to expand his knowledge about a subject? Whom will the user trust?
These questions are just the beginning of the inquiry that a company must pursue to have an effective enterprise search strategy. Marketing and operations have to be involved in the planning stage. They can’t just turn to the IT organization and expect magic to happen. They also have to be involved in the discussion about the user experience. The user experience is what drives behavior to the desired outcome, and a search engine user experience is different than the broader term user interface. Chances are that your user didn’t land at your site’s home page, so the starting point is an indication of the path of travel for your user. Provide signposts for the user that tell him what to do next, where to go, and how to find more information.
A key component of extending the notion of search beyond the home page’s search box is to make faceted browsing available at many points within a website experience. Preventing the user from finding more information sacrifices the potential of broadcasting (allowing the user to expand out inquiry) for the sake of narrowcasting (assuming that the user is looking for one and only one thing). Getting a user to have a “hey, I didn’t know that…” response while on your site encourages inquiry and exploration and builds trust between the user and your site.
Once the users and their experience paths are mapped out, the marketing and operations teams need to determine what the metrics of success are. Broadly speaking, the categories are simple: for marketing, it’s an increase in revenue and for operations, it’s a decrease in cost. Still, at this point, it’s a multivariate analysis, so identify the interim steps that lead to the end result, such as decreases in abandonment and pogo-sticking and increases in time on site or time on a page. Next, measure a baseline. Without a baseline of statistics, it’s impossible to determine what, if any, progress has been made subsequent to a search engine strategy implementation.
Only at that point is an organization truly ready to begin analysis of a search engine technology. Evaluations of license cost, implementation cost, functional capabilities, organizational capabilities, and risk acceptance profiles all play into the determination of what search engine is correct.
Even after implementation of the search engine, the work is not done. Measuring the results, identifying candidates for A/B testing, tweaking relevancy calculations, and reevaluating user personnae mean that search is neither a fire and forget function nor just an IT problem to solve.
I’ve had many discussions with people who tell me that they’ve had problems with [insert name brand search engine here] and now are looking at Solr to improve performance. While Solr will often improve performance and relevancy, the discussion often should be reframed as “we’re not sure what we expect from a search engine, and we need help in determining what we want to do and how Solr can help us to get there.”
Otherwise, enterprises simply are not unlocking the power of a true, holistic enterprise search strategy.
Published at DZone with permission of Jason Hull , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.