Chapter 11: HR Analytics

To explain and demonstrate typical analytical process, CGI Advanced Analytics Teamperformed advanced analysis over anonymous corporate employees’ data. The sample data set represents prepared and clean data integrated from several HR information systems.The dataset contains common and specific HR-oriented features for utmost 1500 individual employees regarding:

Demographics

Job and company environment satisfaction

Travelling

Education and field

Job type, level and status

Time reports and absences

Rates, salary and project utilization For each individual record there is an information determining whether the corresponding employee left the company at the end of analyzed period. This information is used to identify key features connected to attrition issue.Uncovering hidden data patterns to predict present employees in the risk of attrition, outcomes of this study show the way how to save a value by identifying possible causes of talents.

4“Usual approach included in black-box analytical tools is rather problematic. It leads to inaccurate solutions and very incomprehensive conclusions. ”Case-Study Prediction and understanding the attrition of employees To explain and demonstrate typical analytical process, CGI Advanced AnalyticsTeamperformedadvanced analysis over anonymous corporate employees’ data. The sample dataset represents prepared and clean data integrated from several HR information systems. The dataset contains common and specific HR-oriented features for utmost 1500 individual employees regarding:DemographicsJob and company environment satisfactionTravellingEducation and fieldJob type, level and status Time reports and absencesRates, salary and project utilization. For each individual record there is an information determining whether the corresponding employee left the company at the end of analyzed period.This information is used to identify key features connected to attrition issue.Uncovering hidden data patterns to predict present employees in the risk of attrition, outcomes of thisstudy show the way how to save a value by identifying possible causes of talents.

ANALYTICAL APPROACHES The selection and right execution of suitable analytical approach and modelling method is essential for the precision and usability of final outcomes and its utilization.Usual approach included in black-box analytical tools,such as automatic feature selection and choosing from predefined prediction algorithms,is rather problematic.This leads to inaccurate solutions and chiefly, very limited and incomprehensive conclusions.CGI Advanced Analytics methodology brings solid and proven approach defining how to correctly manipulate the data, test and select suitable modelling techniques and discover reliable and highly applicable insights.For the purpose of the case-study, whole dataset was inspected and using logistic regression method, two different questions were answered:Which specific factors increases or decreases the probability of attrition?Which individual employees across different jobs are in high risk of attrition?Regarding the first question, advanced modelling techniques like neural networks or logistic regression are able to identify “drivers” that influence target variable –risk of attrition in this case. In opposite to traditional and trivial methods such as simple correlation, advanced methods are able to get further information and uncover more complex patterns.

The figure bellow shows far most relevant factors influencing attrition:This kind of information provides seemingly straightforward insight. However,in order to deliver more thorough and usable conclusions it is necessary to go a little bit deeper.One possible way how to do that is for example to perform such an analysisseparately for different job types and roles.The next figure shows the importance of various factors for three different job categories: technician, scientist and salesman

In the general meaning,social network analysis term is not only associated to social media like Facebook or Twitter. It also describes a way how to explore various type of social interaction. In the context of HR analytics, this can be used as an instrument how to enhance standard models by adding information about interactions among employees.Real social networks and communication flows in companies often differ from structural organization charts. Social leaders might be identified even they are not on the top of structural hierarchy.

Interactionscan be identified by:

List of calls

Mail communication

Instant messages

Mutual meetings

Work on same project and shared folders access

Answering the second question -who are the employees in a danger of the attrition, prediction model was developed and applied. For evaluating purposes, one third of thedataset was separated to test the model accuracy. The rest was used to train the model and perform previous analysis.Developed model is able to predict 88.9%of employees withleft-the-company flags.