25 02 2014
by Inostix

Why the use of social networks in HR analytics is very promising

Print Friendly

Introduction

networkThe ideas of social network learning have been widely and successfully used in customer churn prediction models in the Telco sector.  The idea here is to predict customer churn not solely based on individual customer characteristics (e.g. socio-demographic information, calling pattern, …) but also based upon the characteristics of the friends/neighbors (e.g. how many friends have recently churned). In this blog article, we would like to explore ways of using the ideas of social network analytics in HR.

 

Important to note here is that a social network is more than just Facebook, Twitter, LinkedIn, etc.  In this article, we will especially look at organizational social networks whereby we  define a network as a collection of nodes (e.g. employees, companies, …) interconnected by means of edges (e.g. working relationship, email link) which can be defined depending upon the setting under investigation (e.g. employee churn, performance management).  Given their promising potential, the interest in using ideas from social networks in HR analytics has gained traction recently.  It is the purpose of this blog to explore some key applications and discuss challenges involved.

Application 1: Predicting employee churn within a specific company

Traditional employee churn models are typically based upon employee specific characteristics such as age, background, department, promotion, performance, etc.  Although this is definitely useful as a start, social network information can help to further enrich the employee data and as such the performance of the analytical models.  A key challenge here is to carefully design the social network in terms of its nodes and links.  Obviously, the nodes will represent the employees.  The links however can be defined in various ways.  Links between employees can be established by means of a family relationship, common employment history, or by means of a common department and/or projects they have been working on.  By including this information into the analytical model, it will become possible to detect more complex employee churn prediction patterns.  

Application 2: Predicting employee absenteeism

Absenteeism can be defined in multiple ways: illness, accidents, etc.  Predicting absenteeism is of big importance to companies nowadays because of the costs and organizational problems involved.  Also here social networks can be very usefully adopted in the analysis.  Consider e.g. the bipartite social network depicted in the figure below.  A biparte network is a network containing two types of nodes, in this case employees and the way they are connected in e.g. teams, project groups, young graduates, functional groups (e.g. ICT developers), high potentials, low performers, etc.  Using these bipartite graphs allows to investigate whether certain teams, project groups, etc. have higher absenteeism rates than others.   Note that it is also important to complement this social network information with other types of information to enrich the data as much as possible.  E.g. location information can also be very useful to predict whether bus drivers cause more accidents on the countryside than in the city center.

Social_Network_Nabour

 

Application 3: Improving employee performance by mining collaboration communities

Social networks can also be used to detect high performing workflow communities within an organization.  Consider e.g. the network depicted in the figure below (taken from http://orgnet.com/email.html)The network represents the email flows amongst various employees in an organization.  Each employee is represented by a node which is colored according to his/her department (e.g. marketing, HR, logistics, …).  The network can be built in a straightforward way based upon email To: and From: fields.  Once het network has been constructed, one can apply community mining to find clusters of employees that frequently collaborate (or do not collaborate but should).  Once those clusters and patterns have been detected, HR (in collaboration with the management) could start thinking of ways to improve the efficiency of the collaboration/communication, review workflows, follow-up on compliance regulations, analyze structural bottle necks, etc..  One interesting option here could be co-location whereby employees that frequently collaborate are physically located together so as to reduce the communication burden and improve their efficiency.

 Email_social_network

 

Challenges with the use of social networks in HR analytics

The above 3 applications clearly demonstrate the potential of social network analysis in HR analytics.  However, 4 main challenges arise for HR leaders when working this out.

  • Defining the network: a first key challenge in doing social network analytics is defining the network.  Usually, the nodes are straightforward to define.  However, the links are less obvious since they depend upon the context of the analysis.  Also, the links may be weighted based upon e.g. intimacy and/or recency of the connection.
  • The impact of the network: once the network has been properly defined, a next challenge is how to take into account its effects in a predictive HR analytical model.  A very popular and powerful approach here is the idea of featurization.  This means that the network characteristics will be added as features or characteristics of an employee. Consider e.g. the example depicted in the table below.  The first three features (age, degree, salary) are referred to as local, employee specific information.  The subsequent two features are network features since they look at the network neighborhood of each employee.  By adding these network characteristics to the HR data set, it becomes possible to analyze it using e.g. logistic regression.  Empirical evidence has shown that including network features in a plain vanilla logistic regression model is a simple and yet powerful way of taking into account social network effects within a particular problem setting.

Screen Shot 2014-02-20 at 17.15.34 

  • Dynamic nature of networks: networks are continuously evolving, e.g. because of employees churning and new employees being hired, employees changing teams or projects, etc..  As such, facilities need to be available to continuously update the networks and learn new emerging patterns.  From an analytical perspective, this not easy to do so however.
  • Privacy: a final major challenge concerns privacy.  E.g.,. think about the example of mining email traffic.  Many companies adopt a policy whereby emails are considered strictly confidential and can as such not be analyzed/mined.  Although this can definitely be defended from a privacy perspective, employees should realize that by giving up some of this privacy, they might get a lot in return in terms of improved efficiency of the collaboration.  In any case, it’s our firm believe that analytical transparency and ethical use of HR data, is the best approach in the long term. Also, an opt-out option for employees should always be available.

About the author:

photo_BBProfessor dr. Bart Baesens holds a master’s degree in Business Engineering (option: Management Informatics) and a PhD in Applied Economic Sciences from KU Leuven University (Belgium). He is currently an associate professor at KU Leuven, and a guest lecturer at the University of Southampton (United Kingdom). He has done extensive research on data mining and its applications. His findings have been published in well-known international journals (e.g. Machine Learning, Management Science, IEEE Transactions on Neural Networks, IEEE Transactions on Knowledge and Data Engineering, IEEE Transactions on Evolutionary Computation, Journal of Machine Learning Research, …) and presented at international top conferences. He is also co-author of the book Credit Risk Management: Basic Concepts, published in 2008. He regularly tutors, advices and provides consulting support to international firms with respect to their data mining, predictive analytics, CRM, and credit risk management policy. In that context, he is academic advisor of the HR Analytics start-up iNostix. Read more on Prof. Baesens on Dataminingapps.

3 Related articles: 7 Benefits of Predictive Retention Modeling and 4 Challenges with Predictive Employee Turnover Analytics and 5 Reasons to start with Predictive Employee Turnover Analysis

Interested in using predictive HR analytics as a key component in your HR strategy? Get in touch with CEO and co-founder Luk Smeyers for more information ([email protected] or via Google+). Or follow iNostix on Twitter and/or Facebook for exciting international articles on HR analytics.

February 25, 2014