Forget about big data. The latest analytics buzzword HR leaders should be tracking is actually a blast from the past: machine learning.
The term has been around since the early 1960s, but over the past two years human resources tech vendors across the industry have infused it with new life. They’ve gleefully announced acquisitions of machine learning analytics companies and made bold predictions about how they will integrate this state-of-the-art technology into their next round of product offerings. Automatic Data Processing Inc., CareerBuilder, SuccessFactors, Ultimate Software Group Inc. and Workday Inc. are just a few of the many companies that have proclaimed a commitment to incorporating machine learning into their road maps, and several of them have already started rolling out products.
Machine learning: The use of algorithms to teach computers how to make predictions based on patterns they find in data sets.
So why does this relatively obscure term matter in the grand scheme of HR? In a perfect world, machine learning will solve the biggest analytics challenge that most HR professionals face — doing the actual analysis, said Holger Mueller, an analyst with Constellation Research Inc.
“An intelligent machine can look at data and make recommendations for you, such as which people should be on which shifts, or who’s at risk of leaving the company,” he said. HR leaders still have to act on these recommendations, but the technology will do all the dirty work.
And the more opportunity the machine learning algorithm has to explore the data, the more accurate the recommendations will be. At least that’s the premise. “We are still at the very cusp of this trend,” Mueller said. And, as with all predictive analytics promises, it is best not to get too excited too soon.
Way Beyond Keywords
It could be years before the benefits of machine learning for talent management are fully achievable, though progress is being made. In 2014, Workday acquired data analytics company Identified and later that year rolled out the Workday Insight Applications, a suite of tools that use machine learning algorithms to analyze talent data and make predictions.
“Each Insight Application will address a specific business scenario or question,” said Amy Wilson, vice president of human capital management products at Workday. The first one looks at who is at risk of leaving the company, and future algorithms will look at things like where to find the best candidates and how to develop high performers.
What’s interesting about machine learning is that unlike keyword searches, this technology can take a more holistic approach, said Rosemary Haefner, chief human resources officer at CareerBuilder, which recently acquired analytics company Textkernel. For example, when reviewing résumés, it can look beyond specific titles and find patterns that suggest a candidate will have the desired skills and abilities. “It delivers a richer set of results so you can find the candidates you might otherwise have missed.”
Then there is the prescriptive part, Wilson said. The machine learning algorithm doesn’t just identify patterns, it provides recommendations on how to react. And those recommendations differ based on the results.
For example, if a person is at risk of leaving a company because of lack of career opportunities, the algorithm can suggest potential promotions, whereas if they are at risk based on low salaries, it might suggest a raise. “It pinpoints the problem then finds a resolution,” she said.
As with all analytics tools, the results that come out are only as good as the information you put in. Workday suggests having at least 18 months’ worth of talent data in place to generate meaningful results — more if you have a small company or low rates of movement among employees.
And, early on at least, HR leaders may still want to do their own general analysis to prove to themselves and stakeholders that the technology works. At Workday, most early adopters of the Insight algorithms are starting small, using them in a contained HR environment and validating results through conversations with managers and their own research.
“Ultimately they want to put this data in front of their leaders, but they aren’t comfortable quite yet,” Wilson said. “This is a good way to shepherd it into the organization.”