… It’s a People Recommendation System that we need!
As the expression goes “beauty is in the eye of the beholder” and this was never more true of expertise. There is no single measurement for expertise since it completely depends on what YOU want from the so-called “expert”. If you are a sales guy looking to close a big Websphere deal, for example, then you are probably not looking for the lead developer on Websphere, but rather you are looking for someone who knows Websphere well but more importantly has had some customer exposure and ideally helped close a similar deal in the past. In this case their “deep expertise” is less important than their “customer expertise”. Now if you are a customer suppport rep and are buried knee deep in a tricky customer problem and need help… well heh! you may indeed be looking for this deep Websphere expert.
Also, a point we frequently forget is that “expert matching” also depends heavily on what THEY want from you (and/or are willing to give you). There is no point in finding someone who is completely unavailable to help you because of other priorities. And from the other side, the last thing you want is for one of your best guys to be bombarded by every Tom, Dick, and Harry in the organization. You need to spread the load across your organization by finding the people who uniquely match the particular need that you have, rather than someone who generically meets your need (ie. is a Websphere expert).
So in a nutshell, to make People Recommenders work you need to ensure a broad spectrum of features are being included in your analysis, features which align to the needs of your organization.