9 Lessons in Enterprise Search Satisfaction

Is “(I Can’t Get No) Satisfaction” the anthem for doomed search applications?

In 2001 I undertook my first major enterprise search project. Search wasn’t initially in the scope of the intranet strategy project I was undertaking for the IMF, but as with so many intranet projects over the years, the importance of an effective search application quickly became obvious.

Last month I shared my lessons learned over the last two decades in a set of Tweets, with a focus on achieved search satisfaction. This column collates these Tweets, with some additional commentary and links to previous columns.

Strategy

Precisely because it is so difficult to make a quantitative business case for search (productivity ROIs are laughable), there has to be specific strategy for search across the organization. Most organizations use multiple search applications and may not even be aware. All of the search applications need to be brought together for strategy and management. However, a search strategy has to fit into a broader information management strategy, as information and content quality significantly impacts search performance and user satisfaction.

Technology

Search technology is complicated. Look no further than Coveo’s 192 page user guide and Funnelback’s documentation to get a sense of this complexity. You need to know enough about the technology to understand both the capabilities (and lack therein) and the extent to which they may support or inhibit any attempts to deliver effective search solutions, especially with personalization.

This is especially the case with natural language processing, artificial intelligence and machine learning. Never take them at face value. You need to understand enough about how they work to assess the applicability of their logic and the validity of the data sets on which they are based, or to suggest and confirm changes. If your vendor is not willing or able to match (for example) the IBM FactSheet initiative structure, you will never optimize search performance.

Related Article: Has Microsoft 365 Been Clinically Tested?

People

Enterprise search applications need the same level of support as other enterprise-wide applications, support which takes into account the number of employees who rely on it every day. The search team should have technical, business, linguistic and user experience skills. The team should act as the hub of a search community of experienced users from across the organization. This begs the question of if organizations know which employees have these skills! This community is an essential sounding board for ideas and problem solving, and also to identify training requirements.

Decision Support

View enterprise search as a decision support application. A focus on tasks is only a partial step towards understanding the information that fuels business critical decisions. Well-informed decisions reduce business risk and accelerate career development, two important justifications for search investment.

When evaluating either a current or potential search application, start with the user interface and work backwards. If facets are going to be important to managing a long list of notionally relevant results, then the index needs to be able to generate these facets.

The design of user interfaces is especially challenging with cross-application federated search. Searching across multiple languages creates similar problems. Combining federated and multi-lingual search is seriously challenging. A vendor may claim to offer 200 connectors, but the question of how the UI will manage these disparate content sources remains. 

Related Article: Is Viva Topics the World’s First Topic Computing Solution?

Performance Assessment

A fundamental challenge of search is it does not support a defined business process. Two employees with the same experience and expertise but in different subsidiaries or countries will query in different ways and use different heuristics to evaluate relevance. An in-depth review will be required here. Set aside a chunk of time to carefully review quantitative (e.g. query logs, phrase use, low and zero results) and qualitative (usability testing, search satisfaction scores) factors as well as the business language (specialist terms, synonyms, acronyms etc) of the organization.

Search Training

Companies invariably train employees on other enterprise applications, even at times requiring certification before giving access to the app. Training on search applications is just as important, because it is about the technology of access as well as the content scope.

Searching is all about interacting with the application. Static views on PowerPoint slides don’t help here. Interactive sessions, conducted by members of the search community are needed. 

Failure Management

Enterprise applications with defined workflows (e.g., adding a new employee to an HR application) tend to make it obvious when the application has failed, if only because there are embedded rules of operation and constant validity checks. Search users have no way of knowing if a query found all the information they are permitted to see. They have to trust the application totally. The excellent failure schematic from Clearbox Consulting is worth familiarizing yourself with.

Related Article: Diagnosing Search Failures

Search Is Not a Project

From the perspective of the search management team, search is more of a vocation than a project. Every day new content is added, new queries are formulated, new employees need training, new security rules arise and new application enhancements need testing before being rolled out.

The Starting Point

Although this column starts off by recommending you develop a search strategy, the most important step towards search satisfaction is the investment in a search team. Search teams can write the strategies, assess the technology, establish and monitor performance, train users, react to failures and plan for the future. No team? To return to the Rolling Stones, “I can’t get no satisfaction” will become the anthem for your doomed application.