What is a proof of causation, really? How can we establish the connection between an investment and a business effect or say which metric drives another? Why should we strive to get to causation using our big data? Because to compete in today’s competitive environment, we need to know which people investments change our business for the better, and how can we improve them.
Ultimately, proof is a combination of mathematics and old-fashioned reasoning and argumentation. It is always possible that factors outside a study — the economy, the weather — may have affected things in ways that were unanticipated. HR analytics is in its infancy, and we are still trying to solve issues that previously couldn’t be measured. Striving to understand how your HR initiatives cause change gets you closer to aligning your investments with your business outcomes.
So to get started, ask yourself the following questions:
• Can you create separate test and control groups for comparison purposes? Is one group receiving the investment (such as training) and another is not? This is a natural formation of a control group since due to budget or classroom sizes, there is usually a group of people who have not yet received training. This group waiting for the intervention is your control group.
• Are there observable differences? Statistics from your data provide objective criteria. Statistics provides methods for assigning a likelihood that a particular outcome was the result of an underlying pattern, rather than just through random variation.
• Did you consider prior performance? This is a crucial component to establishing causation. Any time there is data about the employee’s performance prior to the beginning of a program, it is important to collect that data and determine how the test and control groups differ.
• Did you look at the descriptive statistics? Combining, for example, your HRIS, operational and LMS data gives you an opportunity to look at the relationship between the metrics, training saturation and employee demographics. Powerful insights can be gleaned here.
• Have you considered the relationship between the metrics? When multiple metrics are available, it is necessary to consider the interrelationship between the different values.
These answers will provide you with what needs to be considered before discussing impact on a study. Data is provided to us in many ways and the above factors are produced in hard data sets. However, “softer” data, such as interviews, surveys, success stories and reactions to the investment, will be important for telling a good story and should be consistent with the quantitative data.
Please feel free to comment below with your opinions on causation versus correlation. Next week we will discuss taking big data and getting to business impact.