People and expertise search – the ultimate challenges in precision and recall?

People and expertise search – the ultimate challenges in precision and recall?

by | Jul 6, 2017 | Collaboration, Digital workplace, Search

Surveys of the use of both enterprise search and intranet search invariably show that searching for information about other employees is one of the most important features. In practice searching for a name when you are not sure of the correct spelling, or the person does not use their official ‘HR’ name, can be more than a little difficult. The opening section of my new 16 page report on People and Expertise Search sets out just a few of the challenges that people search has to cope with. After all people search is the best possible example of absolute search precision. A near match is just not good enough!

Over the last few years interest in using technology to find expertise within the organisation has increased significantly, though the basic principles date back to research and development undertaken in the late 1990s. Technological development has been very rapid, partially driven by software designed to be used by police and intelligence services to track terrorist activity. This report provides a summary of the many different approaches to identifying expertise that do not rely on self-completion of personal profiles, an approach that is doomed to failure. Can you write down all that you know? But that begs the question of how we define an expert, and then develop a way of ‘scoring’ their expertise to provide a ranked list. How would you feel if you (and perhaps your colleagues) regarded you as an expert but the scoring model pushed you right down the results list and out of sight. Might that affect your rate of promotion within the organisation and your reputation with colleagues who may be unwilling to trust your judgement?  People with particular expertise may have been working on highly confidential projects and the expertise gained may not show up in a search, or have joined the organisation so recently that years of experience are not reflected in their initial score.

If name search is about known object/precision then expertise search is all about exploratory search/recall. If a search fails to deliver relevant information then in theory you can ask an expert. If you can’t find an expert then there is no other option except to go with what you have and accept the personal and corporate reputational risks. Which is also why ‘partial’ solutions are so dangerous. Is the expert missing because they are not within the scope of the application?

Even when a possible expert has been found they may not be willing or able to share their expertise. There could be language problems or organisational politics.  Expertise search is only going to have a chance of working if there is a good knowledge management culture in the organisation. In writing about the KM aspects of expertise exchange I benefited greatly from discussions with Paul Corney (Knowledgeetal). His recent post on this topic includes six test questions we have jointly developed which form the selection criteria for any expertise search application and its implementation. Paul’s experience in this area dates back to his work at Sopheon in the late 1990s and I have been tracking the research literature for almost as long.

I shared this report recently with one of the most experienced global enterprise search managers I know.  Their reaction was as follows

“I have quite strong views about expertise location systems and generally think it’s impossible to create something at scale that works in the way people imagine it will.  I’ve seen Excel sheets, databases and / or systems for small teams / organisations that can work.  But when you try to scale this up it just doesn’t seem to work in the way people want e.g. they just can’t seem to find the person they want.  Technology, taxonomy, skills, experience, profiles, etc, etc.  It’s a minefield and I must admit one I stay away from!!”

When you read this report you may feel that it highlights the challenges but does not provide many solutions. There are two reasons for this. The first is that the solution has to be specific to the organisation and take into account corporate culture, the willingness of HR and KM managers to be involved along with IT and search managers, and of course the capability of the technology applications. The second is that Paul and I are developing a public workshop on this topic that will take place in London later this year, with a beta version at the JBoye Aarhus conference in November. We can also deliver the workshop in-house. If you are interested in either the public workshop or an in-house version contact either of us. It should go without saying that we are totally vendor independent. Our aim is to provide you with process to implement expertise identification and not to sell you software.

Martin White