The key steps to enterprise search

I previously wrote about some of the barriers faced by many organisations embracing enterprise search.

This Industry Insight will focus on some of the work organisations are considering in implementing enterprise search technology.

There is no magic bullet to increasing search volumes inside the organisation. It's a cart-before-the horse situation. As users obtain more meaningful search results, so they are likely to use the search tool more often. Deliver meaningless results, and the user community will desert you.

It is unlikely that all content and data repositories inside the organisation can be brought into the search sub-system immediately. It is important to bring on board the repositories that can deliver the best search results quickest.

Those with less complex dictionaries are important to think about, and those repositories that are well structured also yield good returns. Enabling those repositories with reach across all employees is likely to increase search volumes quickly.

Two common examples many organisations are considering:

* A people search facility that allows for expected functionality, but that can also help to locate pools of talent in the organisation, if the data is captured properly.
* The ability for users to search all policies and procedures to find a particular page or reference.

Tracking search phrases and matching these to results is critical. Often companies will be required to perform a degree of analysis on the search results to identify commonalities, misspelt words, trends and the like. This input can be incorporated into the company dictionary to help increase the effectiveness of search queries.

There is a need to respect content security. However, certain users can take it too far.

Grant Hodgkinson is MD of Mint.

Different departments are also going to require different dictionaries. Consider a research team and the HR department. Both will have specific phrases and words they use, but those dictionaries are unlikely to intersect in a substantial way.

Carefully analyse the strata of dictionaries in the organisation. Perhaps the requirement is to allow users to hone their search into expertise areas, which can then be linked to a dictionary. Alternatively, it may be appropriate to consider implementing something more generic if the content can be organised in subject taxonomy.

Keeping the company dictionary generic is a sound practice. It could also be appropriate to develop a search-enabled glossary, which could help direct users from one word to another.

Words could have dual meanings. In this way, it might be good to have a drill-down subject hierarchy that could drive search results to users. This concept taxonomy helps users drill down into topics or metadata areas in a navigable way, starting with major topic areas.

Many indexing applications have become pliable in their ability to interact with data repositories and different data formats. Historically, search engines have worked well with HTML and similar content, but all major vendors have made headway in going beyond this form factor.

The programmatic conversion of data from one format to another just for search purposes can represent complexity. This is something to be avoided.

It is sometimes easier to represent content from a foreign repository in a format that can be consumed easily by a search engine. Consider converting or representing this data in a format that can be easily consumed by a search engine.

Security of content and data is a non-negotiable, especially in this era of compliance awareness. In the enterprise search arena, security is not necessarily the permissions on a file share or set of HTML pages. It might be a representation of the users who are allowed to see certain records or columns in a database. The security and authentication system of the content in the foreign repository is sometimes not 100% aligned to the searching sub-system.

Depending on the indexing engine being used, to incorporate security into the content indexing or search process can be a challenge.

The best areas to commence searching would be those where security is not a major consideration, such as anonymous or public domain content.

If security of searched content is going to be complex, then evaluate the search engine carefully, or plan the deployment model thoroughly.

Unless users have jobs that distinctly require it, searching for information can take time. Requiring the user community to embrace the technology whole-heartedly when it impacts their effectiveness in dealing with customers, for example, is unlikely to garner much traction.

The approach can require the creation of hierarchical topic-based searches. Users can expand from one topic to another to find what they are looking for, and results are pushed to users without their having to think of keywords to pull results down. Training and change management will be important.

There is a need to respect content security. However, certain users can take it too far. Here executive support (and active involvement) is critical. How about the executives demonstrating value or pushing people towards a search platform to help drive awareness and usage?

People often consider their strength in the organisation by how much content they control, and have not given over. Changing this requires substantial effort: sometimes coaching and sometimes pushing.

Enterprise search technology can indeed deliver benefits. The road to meaningful results is often a long journey. In this arena, incremental releases of functionality are often likely to work best.

* Grant Hodgkinson is MD of Mint.