Workforce analytics: Avoiding the pitfalls of spurious correlations

By Sheri L. Feinzig, PhD, Director of Strategy and Smarter Workforce Institute, IBM

Sheri L. Feinzig, PhD, Director of Strategy and Smarter Workforce Institute, IBM

The tremendous potential of workforce analytics has been well documented, but most organizations are at a relatively early stage of implementation. As they embark on their workforce analytics journey, organizations face many considerations, from establishing objectives of the analytics program to determining how to turn insights into actions that drive positive change. 

 One area of particular importance is the necessity to truly understand the relationships that are shown within the data. As data continue to explode, the tools for analysis are continuously evolving. This provides a tremendous opportunity for valuable insights to be derived–but also an opportunity to discover spurious or irrelevant relationships in the data. Such findings can potentially lead to poor decisions, misallocation of resources, unrealized benefits, and even legislative violations.  

 When it comes to people data the stakes concerning irrelevant correlations are high, and the spurious nature may not always be obvious. So how can you spot spurious correlations, and avoid drawing the wrong conclusions and making misinformed decisions? There are basic steps that all workforce analysts should follow to stay focused on the things that matter: 

  • Start with Hypotheses 

 Rather than allowing every random relationship to emerge, it’s best to start with a well thought out path to explore, based on logic and previous findings. Your overall approach should be hypothesis driven to allow meaningful relationships to be identified and validated. 

  • Understand Your Data 

 Take a look at the distribution and other basic descriptive information to make sure there are no extreme values (outliers) affecting your analysis. You don’t want some errant data point driving your actions. 

  • Understand HR policies and practices 

 In some cases you may see relationships between data points such as marital status or commuting distance and turnover, but these factors are irrelevant to job performance and could lead to erroneous actions that perpetuate discriminatory practices. You can mitigate this risk by having strong HR expertise as part of your analytics team. 

  • Apply Common Sense 

 Examine your findings, find the golden nuggets, and take actions based on what makes sense for your organization and your people. Don’t unquestioningly follow the data. 

Read Also

Human Capital Analytics Leap Forward

Human Capital Analytics Leap Forward

Rick Lueders, Vice President Business Strategy, Bullseye Engagement
Understanding People Analytics for HR Professionals

Understanding People Analytics for HR Professionals

Dr. Mesut Akdere, Director of Purdue HR Analytics Interactive Learning Lab, Purdue University
Building HR Analytics Capability in Your HR Function

Building HR Analytics Capability in Your HR Function

Linda Croll Howell, Ph.D., Director, Human Resources, Analytics, Cornell University
The Rise of the Skills Economy

The Rise of the Skills Economy

Anna A. Tavis, Ph.D., Clinical Associate Professor, Academic Director, Human Capital Management, Division of the Programs in Business, NYU

Weekly Brief

Top 10 HR Analytics Tech Companies - 2018

HR Analytics Special