CAN DATA MINING BE USED FOR STAFFING AND PERSONNEL SELECTION?
By Ed Colet

Selecting the right personnel to staff a project is one of the more important considerations that will ultimately be responsible for the success (or failure) of the work. In this column, I talk about the role of data mining software to facilitate the selection and staffing of personnel for a project. If it's already used in the sports domains, what would it take to apply it to a business domain?

An example of the use of data mining for personnel selection is the IBM Advanced Scout software application. Advanced Scout is a data mining tool used by coaching staffs of the NBA (National Basketball Association) to look for interesting patterns in the data of their games. One of the key data mining analysis is the line-up analysis. This discovers which particular groups of players are surprisingly effective or ineffective. Based on the evaluations of the circumstances associated with the lineups, the coach can then make strategic and informed decisions about who to play and when.

The obvious question then is that if it's readily useful in the sports domain, can the same "personnel selection" analysis be used in a data mining application developed for a more traditional business domain?

In the sports domain, discovering effective personnel combinations is facilitated by the following: (a) Performance-evaluation metrics are readily available and interpretable. (b) The domain allows the coach to repeatedly observe and test the line-up combinations in practices or in games in order to fully evaluate them. (c) The activities in sports are extremely well defined and the rules of the sports constrain what can and can not happen. Thus situations and circumstances that arise in a game can be planned for. (d) A lot of the relevant data to be analyzed, such as the player's statistics and characteristics is readily available online.

In the business domain, it would be necessary to have similar characteristics if data mining software for staffing were to succeed. Specifically, it would be necessary to define appropriate performance metrics that could be used to evaluate a team or group's performance. Rather than "points" perhaps it could be an ROI-type of measure. In a business setting, unlike the sports domain, it's difficult to evaluate the group's performance in a "practice" session. Instead, much of the activity in business has the spirit of "on the job training". Situations and circumstances of a project can also rapidly change in unanticipated ways, and success or failure depends on the ability to react and respond to new demands that arise. Last but not least, what personnel data would be relevant for mining? Obvious information such as employee skills, qualifications, experiences, current workload, and perhaps even the performance evaluations can be stored in a database to be mined. It's more difficult to store the more intangible information such as their inter-personal skills, etc. in a database. (But NASA as well as airlines have systematically attempted to incorporate these more intangible considerations in their crew scheduling considerations).

Despite some of the above challenges, a data mining application for personnel selection could be promising for a variety of reasons. Currently, in a large scale project with a large pool of employees, a manager or project leader may not be familiar with all the characteristics of employees at his/her disposal, and data mining software could be helpful in assembling a team that will lead to success. It's also possible that without data mining, a manager may be biased by showing some favoritism in assigning employees to projects - limiting the chances for success of the project. Assuming that online data is not biased, then an automated data mining approach would be free from such favoritism. Discovering effective employee groups - ones that no one suspected of putting together could lead to other fruitful collaborations among employees.

Despite the above advantages, data mining for staffing is not widely used. Additional reasons include the following. Data mining is a relatively new technology still gaining widespread acceptance although new applications are slowly emerging. Collecting the data on employees especially "intangibles" that may appear only tangentially relevant touches on the sensitive issues of employee privacy and confidentiality. Staffing is a high-level decision made by a manager, and he/she may generally be reluctant to relinquish some of that role - especially if he/she feels that they know their employees better than any software could. But just as data mining was never intended to replace the NBA coach, data mining in this setting also wouldn't replace the manager - but provide them with a way to make better staffing decisions.

Ed Colet is the Acting Director of Research at Virtual Gold Inc., responsible for developing analytical methods for data mining and for investigating human factors and usability issues of business intelligence systems. At present, he is in the final stage of completing a doctoral dissertation in the Cognition and Perception program at New York University's Department of Psychology. Ed has also worked for IBM Research at the T.J. Watson Research Center. At IBM, Ed was a member of the group that developed Advanced Scout, the data mining application for NBA teams. His research interests focus on statistical methods and human factors.

For more information, see http://www.virtualgold.com.