The field of HR Analytics has come a long way in recent years: not only have more and more organizations recognized the value of data-driven HR, but HR analysts themselves have delved into a wider variety of data sources to provide their insights as well. A great example is the current explosion of using unstructured data, such as text, schedules, and even video, to paint rich pictures of organizational life.
Macro vs. Micro data
Despite this growth, however, there is another source of data that remains largely untapped within HR analytics: macro data. Of course, by “macro data” I don’t mean data you collected using an Excel macro, but rather data that is an aggregation of smaller, individual-level units. This is in contrast to micro data, which is the individual-level units that are being aggregated. Let me illustrate the difference through an example.
Think of micro data as simple building blocks, and macro data as structures created by those blocks, which themselves can be used to make even larger structures. In HR analytics we tend to focus on employee-level data; this makes sense, as a lot of organizational decisions (such as involuntary turnover) come down to the individual level.
That being said, sometimes only analyzing at the employee level can lead to incomplete, and even misleading conclusions.
Macro performance data
Imagine that in your organization, you are looking for low performers to let go. You just take the average of their performance ratings, identify the 20 employees who scored the lowest, and let them go. Simple, right? In reality, sometimes this will be a fine approach, and other times it won’t.
While it could be that the group of 20 you identified are indeed the worst performers, it could also be that the managers for those teams are especially harsh in their ratings, or the work is especially tough in their particular locations: in this way, you could end up letting go of employees who actually are (or have the potential to be) top performers that could have been incredible assets to your organization in the future.
One way to avoid this problem would be to look at the average performance rating for that team, office, or manager, and compare the individual scores to it. People don’t work in a vacuum, and there are many factors that can affect their performance that are unrelated to their true ability; by taking macro data into consideration, you can do a better job of “cutting the fat” and getting to what you really care about.
Macro data in turnover
In an organization, you can systematically gain insights on the individual, team, office, department, and organizational level. Macro data doesn’t stop there, however, because our organizations don’t exist in a vacuum either. One incredible source of data that remains under-utilized in the HR analytics space is the economy.
Let’s imagine your role as an HR analyst is to predict turnover across a large multinational organization. You begin by using individual-level predictors across the whole organization, like salary, tenure, and performance ratings, but you find that your model doesn’t predict very well.
In reality, though, a major contributor to turnover is the strength of the local labor market, and that will change across different offices. The labor market could be booming in Toronto, leading to satisfied employees getting constant messages from recruiters, but struggling in Miami, leading to employees who want to leave having trouble doing so.
In this way, it makes sense to take into consideration something like the unemployment rate associated with different offices in your analyses. Beyond the direct effects, it’s also quite possible that there are interesting “interactions” that can help make your model much better.
For example, you could find that when the unemployment rate is high, tenure is negatively associated with turnover, but when it’s low, the relationship could be negative. If you just run the model overall, those positive and negative effects will just average out, and you will lose an incredible amount of predictive power. In that case, everyone loses except your competitors!
The Short and the Sweet of it
Taking both micro and macro data into consideration can definitely get a bit complex, and sometimes it simply isn’t worth it. But when you really care about getting the right answer to a particular HR analytics question, and you have reason to believe that that there’s some larger factor (a department, a manager, an economy) that’s tying those individual data points together, I recommend digging into that factor: when you mine a newly discovered area, you might just hit gold!
A Bonus for Stats Nerds
In this post, I have described macro-level data on a fairly holistic level in order to simply raise awareness of its utility. If you have a strong background in regression and statistics, and would like to learn more about how to different levels of data (macro and micro) can be represented together in one analysis, I recommend reading on “Multilevel Modelling”/ “Cross-level Moderation” here and trying out the “multilevel” package in R.
What these analyses essentially do is statistically isolate the effects of macro-level factors from micro ones by 1) taking group averages into consideration, and 2) representing residuals (errors) in a different and more appropriate way given the nested structure of the data. That said, these analyses can get quite complex, and it’s often fine for an analyst to just look at the macro data separately.