Last fall, an article appeared in the Harvard Business Review called “Competing on Talent Analytics.” While most businesses turn to analytics to refine business processes, the article details how a handful of
organizations are also performing analytics on employee data to help invest in their human capital.
The authors -- Thomas H. Davenport, Jeanne Harris and Jeremy Shapiro -- describe the field as relatively new, but, like general business intelligence (BI) best practices, businesses should begin with good data, start out basic and work their way up to more advanced, more sophisticated approaches.
SearchBusinessAnalytics.com recently discussed the field of talent analytics, how businesses should get started and what pitfalls to avoid with Davenport, the president’s distinguished professor of information technology and management at Babson College in Babson Park, Mass.
Your piece implies that employees should be thought of as capital. Why is that?
Thomas H. Davenport: Among the more sophisticated companies, it’s a pretty well-known idea, and it’s been talked about by academics for a really long time. Even Adam Smith defined it. Companies started to pursue it seriously in the 1980s and 1990s and so on. And there’s a Nobel Prize-winning economist, Gary Becker, who wrote a book in 1964 on human capital … The benefit is that you maximize their value to organization. If you think about them as investment, you will think about how to get the most out of your investment and will figure out the best level of investment to pursue. Organizations that view their employees as human capital tend to treat them better and worry more about hiring high-quality people in the first place, and they tend to focus more on training, education, knowledge and expertise.
What is “talent analytics?”
Davenport: It’s the idea of using data and systematic analysis to get the most out of your people. I think, like all analytical disciplines, talent analytics requires that a fair amount of data be in place to do much of it. ERP systems and HR [human resources] information systems, in general, have a lot of data about people now in the usual pattern, replicated in this space. First, organizations pursue the transactional capabilities, then several years down the road, they’ll say, “We have a lot of data here maybe we could use it to make decisions.” … HR analytics, for most organizations, is one of the last [analytics areas] to pursue.
Why is HR one of the last places for analytics?
Davenport: Partly it’s a matter of data. Human resources is one of the last areas to get transactional data and to get good information about employees. Partly it’s that HR is historically not that analytical or computationally focused. Instead, it’s more of a people-oriented discipline.
Examples in your Harvard Business Review piece reflect that larger organizations -- Google, JetBlue, Sysco -- are performing talent analytics. Would this also be appropriate for midsized and even smaller businesses?
Davenport: I think it’s certainly true that large organizations are the first to pursue it, but that seems to be case for analytics in general. If you have a smaller organization, there might be more face-to-face interaction and an understanding of people’s capabilities, but I still think it’s valuable to predict, for example, who will be a high-performing employee and who is likely to leave the organization. That might be difficult with 50 employees, but it’s certainly possible with 500.
How should businesses get started?
Davenport: The first step is having good human capital facts, facts about employees. That certainly includes their level of performance. Most organizations have some form of assessment of their people, certainly middle-sized organizations. Some form of an HR assessment is basic these days. The data includes things like tenure, level of education and various other demographic attributes of employees most organizations have.
To start to get predictive, though -- to predict who might be good employees or who might be likely to leave -- you typically have to gather what’s called bio data, a set of detailed facts of a sample of employees so that you can run lots of statistical relationships between the bio data and performance. You wouldn’t gather this kind of data in the transactional HR system. A great example of this is at Google. The bio data found the best predicators to becoming a high-performing employee is if you’d set a world or international record or you started a nonprofit organization. If you gather a lot of different background factors about a sample of employees, then you can start doing correlations and start thinking about what’s related to high performance.
So, businesses should aim at eventually scaling up to perform predictive talent analytics?
Davenport: Most organizations with any sort of BI spend a lot of time on reporting. And you know, it’s useful, but it’s only about the past. It doesn’t tell you why it happens. Becoming more predictive and statistical helps predict what might happen in the future, and businesses will have a better understanding of why things happen. Moving from reporting to predictive is a good idea … [But] you don’t want to get ahead of your decision maker’s orientation in this regard. If you’re having a problem understanding reports, you don’t want to do predictive work.
What pitfalls should businesses be aware of before jumping into talent analytics?
Davenport: It’s always dangerous to only rely on numbers and quantitative analysis in terms of hiring and firing and so on. Certainly there’s a strong human dimension to talent management. I worked at one point at a professional services firm that decided they could take job applications for administrative assistants, score them on a basis of desirability and never need to be interview. Just hire based on a score. I don’t know if the program was considered a success, but I’d never heard so many complaints in my life. You don’t want analytics to be the only source of judgment about people. There are still plenty of things that you need to talk to people about.
Also, not being very transparent about how the data is being used and the reason why they’re going through this analysis. People tend to be understandably concerned when a business wants to gather pieces of information about them to be used for analysis. Be quite transparent about data collection, what analysis you’re going to use and why you’re going to use it with any sort of analytical approach.
You don’t want to get out ahead of your decision makers, so the analytics are intended to inform decisions. If you haven’t talked to your managers about how to use the program and if you don’t have the buy-in from them, you’ll probably be spectacularly unsuccessful.