My heart sinks when I see an email arrive that starts “We are about to start a search project and we need some assistance. Can you help?” At least with an email I have an opportunity to prepare a creative reply. When the telephone rings with the same question, then the challenge of delivering sad news in a kindly and constructive way is significantly greater. The sad news is that all enterprise search projects fail. Here are just a few of the reasons why.

Project Plans are Built on a Zero Information Base

Project plans involve defining work packages, deliverables, risks and milestones. In the case of most companies no search project has ever been attempted before and so there is no corporate knowledge of what the work packages might be and what skills and resources might be needed to undertake the project.

Usually there is recent corporate memory about CMS implementation or ERP migration projects but these have little bearing on the design of a search project. The end result is that the project plan is no more than a work of inspiration and fiction. 

Undeliverable Deliverables

Project managers spend a lot of time writing a Project Initiation Document (PID) that sets out what will be delivered. In the case of a search project that is often about “delivering improved search satisfaction” or “enhancing search performance.”

Both of these are very desirable goals but no one (least of all the project manager) has any idea of what the current level of satisfaction is with search and what is a reasonable metric for improvement.

The idiom “if you can’t measure it you can’t manage it” applies to search as much as to anything else. If the project is IT-driven then there is a danger of focusing on technical performance (server utilization) or at best increasing the number of searches, but neither will generate long-term benefits.

Content Quality is Not in Scope

Without good content quality no search engine is going to deliver quality (trustworthy) results. If you search for Marmite on the Unilever UK public website, two of the results in the first ten are duplicates of two others and three of the titles are Microsoft Word -- Document 1.

However working on content quality improvement (and nothing improves search more quickly!) is usually out of scope as no one is every quite sure who is responsible for content quality or even for defining what is in scope to be crawled and indexed.

Search Does not Exist in Isolation

Search interfaces with a great many enterprise applications. Websites, SharePoint, document management, social media, business intelligence... the list is endless. Fixing search without understanding how search in other applications is being managed is a waste of effort.