SELECT PERSPECTIVES BLOG

The 6 Do's and Don'ts of HR Analytics

Posted by  Greg Kedenburg

HR-analytics.jpgHuman resources analytics is a very popular topic right now. The opportunity to learn more about your employees/applicants based on data you likely already have can be an HR professional's dream come true. When done properly, analysis of your company’s HR data can yield highly valuable information related to topics ranging from employee engagement and satisfaction to turnover and counterproductive behaviors.

If you know both what you’re looking for, and how to look for it, you can uncover extremely useful data about your employees, which you can then use to improve their experience at work and organizational performance overall. However, the process of investigating HR data can be tricky, there are various pitfalls when it comes to the actual analysis piece of the puzzle. To this end, we’ve put together some general recommendations and guidelines to help keep anyone interested in utilizing their own HR data on the right track.

1. Choosing an Analytical Strategy

When it comes time to analyze HR data, you’ll first need to determine which specific analyses and/or formulas are best suited to answer the questions you are interested in.

Do: Make sure to do your homework ahead of time to figure out which types of studies and analyses are most appropriate for what you’re trying to accomplish. Revise your strategy at any point if you realize you aren’t getting accurate data or data that is helpful to answer the questions of interest.

Don’t: Use a half-baked approach due to not having enough resources. Don’t stubbornly stick to your original approach if you determine it isn’t working, it’s better to have no results than to rely on incorrect results.

2. Statistical Power and Sample Size

In statistics, power refers to the likelihood that an effect or trend in the data will be noticed. The more power a study has, the better chance an effect will be noticed if there is indeed an effect present. Sample size, or the number of data points in the study, is one way to obtain sufficient power.

Do: Make sure you have a sufficient sample size for the type of analysis you’re planning to run. You can run a power analysis in order to determine if you have adequate power or to decide how large your sample needs to be to achieve a desired level of power. Plan ahead and allocate the necessary resources needed to achieve the desired sample size to ensure the analysis has sufficient power.

Don’t: Run blindly into the study without planning ahead, or attempt to run any analyses with insufficient data. After your study is complete, don’t extrapolate any results if you have determined your power was too low.

3. Correlation ≠ Causation

A correlation is a statistical technique which determines if two variables are related. A correlation can be positive or negative. A common mantra in statistics, correlation does not equal causation, refers to the idea that just because you notice an association or trend between two variables it does not mean that a change in one variable causes the change in another. For example, say there is a correlation between ice cream sales and shark attacks. Clearly this does not mean that eating more ice cream causes shark attacks, but rather is due to an outside factor which in this case is the time of year (i.e., summer).

Do: Understand that any correlation you observe may be random and that just because you find a significant correlation does not mean that one variable is causing the other. Exploring other explanations and factors that may be responsible for a relationship appearing will help you maintain an unbiased attitude and objectivity.

Don’t: Automatically assume you’ve found the cause of your problem if you see a significant correlation. Don’t immediately pass along any information implying a causal relationship until you can confirm that the association is most likely not based on chance or other outside factors.

4. Sample Representativeness

In statistics, a population refers to the overall group of people you’re interested in, and a sample refers to the subset of individuals from that population that will actually be taking part in the study.

Do: Ensure that your sample is representative of your population. In other words, your sample’s demographic data, be it gender, race, etc., should mimic the demographic breakdown of the population.

Don’t: Allow your sample to contain overrepresented groups. Inaccurate sampling will lead to a sample that does not accurately reflect the overall population from which it was drawn, which will limit your ability to generalize your results.

5. Making Generalizations

Generalizing in statistics refers to the extrapolation of an effect seen in your sample to the population that the sample came from, or generalizing across time, setting, culture, etc.

Do: Limit your conclusions to your population and other similar populations. In other words, be clear that your results may only hold true in other similar settings or populations.

Don’t: Assume that any effect seen in your study will automatically extend to all organizations, positions, industries, cultures, etc. For example, just because you find a relationship between extraversion and job performance in a customer service position does not mean that such a relationship will hold true in a manufacturing environment. Don’t over generalize your results to other populations that are different from your own.

6. Practical Significance ≠ Statistical Significance

Statistical significance refers to whether a result is due to a genuine effect rather than random chance while practical significance refers to whether a result can be useful in the real world.

Do: Understand that a statistically significant result does not equate to practical significance. Have a critical eye for statistical results and be discerning when it comes to implementing any changes based on these results.

Don’t: Attempt to turn every statistically significant effect into a practical effect. Just because you see a corollary between months that have an eclipse and employee productivity does not necessarily mean there’s a way to leverage that effect into a useful business practice.

Haven't had enough? Click here to learn about 5 more do's and don'ts of HR analytics!

Conclusion

Human Resources analytics can be an extremely helpful tool to have at your disposal. It may seem daunting to make sure you stay within the guidelines listed above, but making sure you adhere to the highest standards when analyzing your data will ensure that your conclusions are accurate. If you’re planning on undertaking an HR analytics project, squaring away all of the details up front according to these recommendations will help minimize any wasting of resources, and will set you up for success in the end.

Also contributing to this blog article are: Trevor McGlochlin, Research Analyst at Select International and Alli Besl, Research Consultant at Select International.

Learn more about sales focused assessments.

Tags:   big data

Greg Kedenburg

Greg is an I/O Psychologist living and working in Chicago, IL.

Subscribe to Email Updates

Recent Posts

Tags

see all

Discover the cost-saving benefits of hiring the right employees, the first time.

REQUEST A DEMO