Here’s a short piece about the practice of adding “best bets” to enterprise search (“best bets” are essentially ponters introduced at the top of search results that direct users to documents that should have the highest ranking).

This article is based mainly on Dennis Deacon’s comment about setting up “best bets” in enterprise search on his own blog, where he points out that instituting “best bets” is often a workaround that’s undertaken after paying thousands (if not tens or hundreds of thousands) of dollars to implement enterprise search.

The first article also points out that adding “best bets” into any search application is a always going to be a manual hack, that is essentially a workaround for fixing what’s wrong with your enterprise search installation.

But what exactly is the real reason for having to institute these “best bets” in the first place? The answer is actually quite simple: poor metadata.

As Dennis points out at the end of his post, “best bets” should never prop up poor search results — yet underlying this statement is the implication that it’s often better to work at improving your metadata over the long term than it is to mask the problem with best bets.

Inaccurate, non-existent or improperly-interpreted (from the engine’s side of things) metadata is the root of problems with search results. Yet fixing problems with metadata (especially in a large organization) usually means one extremely important thing: clearly articulating the link between accurate metadata and better search results to content owners/managers.

If there’s one thing content owners care about, it’s that their content gets found. These same content owners often employ or have relationships with the people that actually post content — and these content posters know how important the content owners believe their content to be.

Yet when content owners ask for “best bets” in search results, what they are really asking for is to have their content highlighted, and given the importance they think it deserves. Now I’m not suggesting that their content is not important, merely that there are better ways to make sure it gets noticed.

Any avid blogger, journalist or search engine optimization specialist knows the importance of the title of a piece. Being vague or misrepresentative in your title will not only hurt your search engine rankins, it will drive your readers away as well.

This is all to say that educating the content owners and managers as to proper metadata creation is often not enough — depending on the size of the organization, it will often take a centralized group who monitors and controls metadata to ensure that users are able to find anything in search.

It’s really just a question of scale. In a smaller organization, it may be more feasible to simply educate everybody out there who’s creating content — while in a larger organization, you’re really going to need that centralized group that monitors and fixes metadata for you.

Unfortunately, I think we’re a long way away from removing humans from the search relevance equation. Surrounding any type of metadata/search implementation is the need for someone to look at things and make decisions about them — a job that technology is currently not very good at replicating.

This doesn’t mean that things won’t change in the near future. But until then, it’s up to your organization to find the human-technology balance that works for you — and it’s more a process of trial and error than a prescribed formula.