As with many realms of life, the human resources discipline has access to more data than ever to inform decision-making processes. It’s now common for HR professionals to employ predictive models based on big data to identify, develop and retain employees. Data on employee attributes can be correlated with past success and used to predict future performance. Unconscious biases and prejudice can be eliminated. Soon we won’t even need humans in the HR department, right?
Not quite. While the unprecedented access to data and analytical tools will certainly help the HR profession to advance by leaps and bounds, it’s not a cure-all. HR teams should use data to inform decision making, but they shouldn’t let data make decisions. Here’s why:
Beware of black swans
People analytics is great for using past data to model future performance – as long as conditions remain the same. Once the fundamental assumptions of the model change, however, the data’s predictive power is lost. In the financial markets, these phenomena are called Black Swan Events, a concept based on an ancient Roman saying that used the term black swan to refer to something that didn’t exist. Of course, we now know that they do.
Industries are being disrupted all around us, and when this occurs our historical data is often of little use. Previously, an engineering background might have been the greatest predictor of success at an auto company. As cars get “smarter”, a background in computer science might be more positively correlated with employee performance. When a shift like this occurs, the predictive models need to be updated as well.
Don’t overlook thick data
When tech ethnographer Tricia Wang worked at Nokia during the early days of the iPhone, she lived with migrant workers in China and concluded that low-income consumers would do anything to get their hands on sophisticated smartphones. She presented her findings to Nokia, but they were dismissed because of her small sample size and because Nokia couldn’t see this trend in any of their existing data sets. We all know what happened to Nokia after that.
This anecdote reveals that big data is often missing the element of human narrative that can emerge from qualitative data, which Wang calls “thick data”. Ideally, a company uses a blend of thick data and big data to get a complete picture of human behavior. People analytics is no exception.
Sometimes data alone isn’t enough
In 2012 Google launched a study to determine what characteristics high performing teams had in common. After studying 180 different teams, they weren’t able to find any clear patterns. On some occasions, two teams would have nearly identical characteristics and overlapping membership, yet one would be successful and one a complete failure. Google was only able to get to the heart of the issue by integrating learnings from the field of psychology.
As it turned out, successful teams were the ones who were able to create “psychological safety,” or an environment where team members felt free to experiment and test out ideas without fear of criticism or punishment. This is a great example of data and people working together. Without a human to apply academic theory to make sense of seemingly chaotic data, Google might never have figured out what makes their highest-performing teams tick.
People analytics has been great for the HR profession, helping us to eliminate unconscious biases and accelerate workforce productivity. But we need to remember that the “human” in “human resources” remains important too.