I was reading an internal blog article by Greg Lang around how we can better re-use internal IP in the services organization. To summarize it quite extremely it boiled down to ranking and and wiki like collaboration.

In a way that is not bad but I feel that if you take that approach you miss a few important items: Information that never gets submitted to the system, Metadata and Search/Action Analytics.

Data that never gets submitted/entered into KM system will always remain elusive to find. So the way to fix that is to make submitting data to the KM system automatic and integral to daily routines. Microsoft's Knowledge Network for MOSS 2007 does that through analyzing peoples email. Email tends to be a very good indicator of peoples knowledge. But of course instead of the link to the people you would like access to the actual knowledge. You want access to the architecture document on search that your colleague across the country has written for a similar size customer. (We will talk about the privacy aspects a different time; but don't worry as they are addressable). Basically you need to find a way to get all data in the enterprise into your knowledge management system not just the docs that people explicitly submitted.

Then there is the issue of lack of good metadata. The better the metadata is that is attached to a document the better the search results will be. Imagine I am looking for a best practice document that describes an enterprise implementation of Search using Microsoft technology (and possibly some partner software). I would go into our IP management system and query for "search + enterprise + architecture". I just did it to get a feel for the quality of the results and it was depressing. most documents are low quality, high level sales pitches or generic portal documents. I couldn't find a single high quality document about this topic and I can not believe that we haven't done engagements about this. The problem was that lots of document showed up but they didn't have proper metadata attached. It turned out that the document that I thought to be the most promising was a level 200 sales pitch. If it would have had attached metadata attributes for level (200) and created by role (marketing) I would have known up front that the document was useless for my purpose.

To takes this a step further I feel that Office Content types allow for much more targeted creation of information in a much more structured way. Content types inherently contain a lot of metadata that could be put to good use by search engines.

Then there is the issue of Search/Action analytics. Here I have to use the example of how good Amazon deals with this:

1. if you look for something very specific like a CD album it will provide you feedback on what customers who liked this album also liked (great to give you new ideas)
2. if you look for a generic item (e.g. HDTV) and select one of the search results you get the follow extra information
1. what do customers ultimately buy after viewing this item (a few choices are listed with their percentage)
2. customers also bought (useful to find accessories that you might need)
3. Many items have reviews available for them
1. there are ratings on reviews
2. you can find all the reviews of a reviewer you like
3. you can provide comments on ratings

Imagine you can do all this analytics on the knowledge that hangs around in your enterprise.

I think that the three topics I talked about make up an important part of a KM system in an enterprise. Of course it is not all of it and I will try to come back to this topic in the next few weeks to cover some of the other aspects.