So I was reading at Bex‘s post last week on Why Google Will Never be Good at Enterprise Search, and its great comments. I ended-up reading several posts out there in the blog sphere on the topic.  Search has been creeping up more and more in my daily work and I figure it isn’t a coincidence as trying to grab stuff from legacy systems or from multiple silos is challenging.  Heck, just trying to find things that some colleague created last year can be tricky.

The Problem with Google Search Appliance

Bex covers this beautifully in his post, and in his previous post on Search Engine Optimization for the Enterprise. Essentially, the problem with Google is multi-fold:

1. When I look for something in the web, there are lots of places with the answers.  When I look for something at work, it is usually only in one document. Precision is more important. Bex describes Internet search well, People aren’t looking for THE answer, they are looking for AN answer. There are lots of answers out there.
2. When I search on Google, if I don’t find what I want on the first page I read, I can either go back to the search results OR follow a link on the page. Wikipedia pages are loaded with links to other pages and to source pages.  At work, everything I look at has no links.  Most documents in repositories, regardless of format, are devoid of links.  Historically the underlying systems don’t support linking from within the documents. You can add the link in Word, but it breaks when you put one, or both, documents in a repository.
3. No links? Sure makes ranking relevance is tough. That shoots the core basis of the algorithm.
4. There is a ton of content in the Internet and Google presents it the same, regardless of who you are. If my boss and I search for something, we’ll get the same hits.  At work, I can only search systems that I can authenticate against and have authorization to access. When my boss and I search for something, our hits are not the same. When you consider that each system may have it’s own Identity Management approach, which is another problem entirely, you can see complications start to pile-up.

Now, in Google’s defense, the existing engines for the Enterprise aren’t a panacea.  They just have a head start in solving the Enterprise search problem.  Even a team of FAST and EMC engineers took time just to get the FAST engine to correctly, and efficiently, index a Documentum Repository.  they had all the expertise and a well-defined scope.  One of my clients has a multi-node index in D6 that now hums.  When the engine first came out in 5.3, FAST search hummed for no person outside of a development environment.

Let’s Get Social

Daniel Tunkelang noted some of the same issues to making search a utility and proposed two approaches to making Enterprise Search work.  With Enterprise 2.0 changing the landscape, we can actually choose what’s behind door number three.

Enterprise 2.0 isn’t just a fancy term or interface.  It is an approach to working that not only helps the Google Search Appliance, but can help traditional enterprise search engines.  The post that inspired Bex, Social Search Wins, and its follow-up, explore how these Web 2.0 concepts can help users find things in their environment by grading results. This goes beyond the obvious implementation of a wiki pages for content that link to each other.

More importantly with Social Search, if we don’t find the answer, we can find an expert that CAN provide the answer.

Identifying users that provide answers and content is the first step. The second is determining the quality of the answers.  If I give out lots of answers, it doesn’t mean that I know what I am talking about regarding any topic.  Just look at this post and you’ll see that I’m not a search expert. It just makes me someone that is likely to be willing to help.  If this post is highly rated and linked to from discussions, wiki pages or blogs, then that may make me more of an expert.

This applies inside the firewall just as well, except the environment is limited to a finite number of systems.

The last stage is the social networking aspect.  If a “friend” links to an answer, or rated an answer highly, you are more likely to want to see that answer first. It isn’t just a trusted source, though that is a significant consideration.  The crux of the matter is that people that are close to you in the social network are most likely looking at that topic with the same perspective.  That adds an even greater strength to the reference.

This approach doesn’t solve all the search issues.  Some answers are in databases or require a combination of data and content.  That is were mash-ups can come into play or search engines can learn to integrate the results from both.

Heck, if I can just find a copy of that proposal we wrote two years ago that had the write-up I want to use now, that would be great.  Lucky for me, I know who to ask. If we were 10 times bigger, I may not be so lucky.