A frequently cited selling point of intranets and social networks is the ability to find people with specific expertise.
The theme came up frequently during three JBoye Group Meetings I took part in earlier this month in Germany. Businesses clearly see finding expertise as a critical task, but fail to appreciate the complexities involved in meeting the requirement.
No Lack of Research, But Challenges Persist
A classic paper from IBM offers some important insights into people search within the company. The usual tactic involves asking employees to complete some form of profile page which explains their background and accomplishments as the basis for expertise search.
This approach fails to take into account the observations made some years ago by David Snowden:
- Knowledge can only be volunteered, it cannot be conscripted
- We only know what we know when we need to know it
- The way we know things is not the way we report we know things
- We always know more than we can say, and we will always say more than we can write down.
Other resources, such as the 2012 book "Expertise Retrieval" and a thoughtful research paper by Professor Morten Hertzum which reviews insights from over 70 research papers, give an indication of the scale of the research into expertise search and the challenges of a successful implementation. Chapter 11 in my book Enterprise Search seeks to provide a practical introduction to expertise search. Volkswagen also offers a good case study and IBM released a further paper on the topic.
Testing Expertise Search
Although most organizations do some form of document search testing (as far as resources allow), from the conversations in Germany it appears no rigorous testing of expertise search, or even people search takes place. Several of the people demonstrating intranets searched for their own name, but did not try to find the profile of someone whose name (Christian vs. Kristian) may have more than one possible spelling.
The first requirement when testing expertise search is to locate specific areas where finding expertise would be crucial to business success, for example someone skilled in the design of bearings for drills working at low temperatures.
Note, expertise search will invariably result in a three or four word query with an implied AND which may inadvertently exclude people who only match three of the skills. The refiners used for document search rarely work well with expertise, so businesses would benefit from having some form of taxonomy or thesaurus for expertise search. In the example above what is meant by the word ‘design’ and how low is ‘low’?
After gathering a set of profiles, there are two actions to take. First do a blind test with (in this case) the VP Engineering and ask them to list out who they regard as the experts. Comparing this list to the search results can be very instructive. Second, ask the experts whether they think the list is complete.
In my experience, the search results raise very important questions about the performance of expertise search.
Onboarding New Expertise
Organizations hire new employees specifically for the skills they'll bring to the company. Some are turning to social media as a way of identifying expertise, and again IBM has been in the forefront of understanding the possibilities here.
A social media-driven form of peer review can provide value, but misses out on new employees who may take a year or more to be widely recognized from social voting.
Previous employers may place restrictions on employees sharing their work after moving on, particularly if the new employer is a competitor. This too will impact expertise search among new employees. Even within the same organization, client confidentiality may restrict the level of detail that someone can disclose.
There is little point in hiring highly-qualified people if their skills and experience are — to all effect — invisible for even just six months.
Total Expertise Search
Sinequa has pioneered an approach that undertakes a federated search across internal and external databases and integrates the results. Astra Zeneca is a good example. HP announced a similar approach with its Collective Project in 2012, with one manager citing the issue of new joiners as a major concern, but it appears to have since vanished. Of course Delve and other graph-based applications also have emerged as expertise search applications.
With all of these releases, one thing remains true: organizations must perform rigorous testing of their expertise search if employees are to trust the results.