“Some people think emotionally more often than they think politically. Some think politically more often than they think rationally. Others never think rationally about anything at all.”
When he wrote these words in Space Chronicles, astrophysicist Neil deGrasse Tyson was referring to the driving forces behind space exploration advances. But the above quote also applies to the world of HR, a field that deals with “agents of choice” who are both rational and irrational.
Perhaps this is why “predictive” has become a strong buzz word in HR circles: Wouldn’t it be nice to know with 100 percent certainty how people — who make both rational and irrational choices — will behave in the future?
And while absolute certainty is never possible, predictive analytics can help organizations look at past workforce behavior to determine what is most likely to happen and plan accordingly. But watch out: predictive analytics is no panacea, and you will be in a better position to take advantage of this emerging technology capability if you can sort fact from fiction. Here are three common myths to look out for:
Myth: Predictive Analytics Will Replace Human Intervention
Fact: Predictive analytics won’t tell you the one clear course of action, particularly when dealing with talent decisions.
Consider what happened when Google came up with a formula for promotion decisions. The algorithm proved to be 90 per cent accurate, but it was never used. As Prasad Setty (Google’s VP of People Analytics) describes in this talk he gave at the company’s re:Work conference in October 2014, the algorithm wasn’t adopted because it was missing one crucial factor:
As Setty describes here, “people need to make people decisions.” The role of analytics is not to replace decision makers with algorithms. I always coach that analytics and data represent evidence, not proof, and it is this evidence that can make our decisions better.
What predictive models can do is focus, inform and guide managers and HR leaders to better decisions in areas that are critical to the business. If the analytics show that a star performer is at risk of resigning, for example, a manager can use that to guide what skip-level one-on-one meetings the person needs for career advancement.
With predictive analytics, we can overcome the limitations of the human mind. As Bruce Hood, a psychologist at the University of Bristol, writes in The Self Illusion, human reasoning “in terms of probable outcomes is very difficult because most of us think in a very self-centered way.” Accurate, predictive algorithms can increase the likelihood that an optimal decision will be made in a timely fashion. However, understanding the limits and benefits of predictive models is key to their effective use.
Myth: The Datification of HR Starts With Predictive Analytics
Fact: Analytics is a journey.
Many HR analysts have shared with me that their manager or a senior executive has stated that “we need some more predictive analytics.” I liken this to trying to drive race cars before learning to walk.
To achieve the value of predictive analytics, some foundational elements must be in place first. Analyst firm Bersin by Deloitte illustrates this well with its talent analytics maturity model, which outlines the different stages an organization typically goes through before working with predictive models. Most organizations start the journey with operational reporting, with analytics teams responding to requests from managers and business leaders wanting to identify problems or understand trends. As the analytics maturity level improves, organizations progress from operational reporting to strategic and predictive analysis.
The case for an in-depth maturity model
We advocate for a slightly more in-depth maturity model that includes three areas that must be balanced: the breadth and quality of data, the ability to support broad and diverse users, and finally the depth of insights that can be produced — of which predictive capabilities represent the highest level of maturity.
The reason this framework is important relates to what we have seen time and time again in organizations that are successful with predictive workforce analytics. They have taken an approach where Industrial Organizational Psychologists, or other statistics-heavy specialists, will do research on potentially high business impact topics. However, to actually turn insights into outcomes, they require access to the right data. The most insightful analysis of low quality data will simply lead to the wrong conclusions, and it isn’t possible to perform analysis, when you don’t have the data to begin with. Further, discovering the insights is only half of the work, as those insights need to be turned into a story to inform and engage stakeholders, and then operationalized – which requires the ability to securely share information to executives, HR business partners, line managers and other potential stakeholders on a regular basis.
Myth: Predictive Analytics is All About Predicting the Future
Fact: Predictive techniques will reveal more than just what is likely to happen.
Ask someone to picture predictive analytics in their mind, and what they will invariably imagine is a trend line that continues into the future. Indeed, the term predictive has become synonymous with a broad set of techniques that can help find patterns and make connections in data, and yes, make a potential prediction about a future outcome — but not everything about the future is really predictive analytics.
The difference between data-mining and predictions
One of the first areas we pioneered using predictive techniques was in the area of retention. While applying predictive techniques to retention has been a long sought after goal, we felt previous attempts were either too simple or gave answers that were not actionable. To address these shortcomings, we first focused on trying to unlock the hidden insights that could determine why people were leaving the organization.
While we also provided a risk score for who is likely to leave — a prediction — we were aware that some of the most valuable insight actually comes from determining the drivers of attrition. (For example, promotion wait time versus promotion readiness, how connected people are to the organization based on the frequency of management change, or more simple aspects such as tenure.) This is not technically a prediction, and is usually referenced as a type of data mining, which is all about finding patterns and trends in data. From an understanding of why people are leaving, it is far easier to go after root cause issues and tackle issues like retention holistically. This may have more impact than trying to gain an accurate picture of exactly who will leave the organization at a future date.
That being said, there are cases where predictions are necessary: One of the transformations we recommend every HR organization make is to build holistic workforce plans that balance the need for talent with the cost of talent. When you have done so, you have created an alignment with the business on what talent is needed, how they impact the business goals, when they are needed, and what it will cost to make this investment. From this, you gain the ability to measure whether or not the organization is on track. The best way to determine whether you are on track is to do something known as forecasting, which involves taking your current progress – in hiring, retention, costs – and projecting those out to determine whether goals are likely to be met. So, you are making a “prediction” about your ability to make your goals.
Making Decisions at the Speed of People
So the next time someone within your organization starts asking about predictive analytics, there is no need to get anxious. It’s important to keep in mind that you won’t get these capabilities overnight, that analytics require human intervention, and that fortunetelling isn’t always the best course of action. If you understand the limitations of predictive models, you will be in a better position to make effective decisions about people, who are both rational and irrational agents of choice.
About the author: Dave Weisbeck
Dave enjoys problems that require both logical and creative solutions, and thus exercise both his left- and right-brain. He started out his career in the 90s writing code as a computer programmer, and then moved on to product management, marketing and general management roles. Dave has a strong background in analytics, having played a key role in the analytics businesses at SAP, Business Objects, and Crystal Decisions. At Visier, he looks after product and market strategy. A proficient do-it-yourselfer (he made his own PVR for fun), Dave’s hobbies include the logical and creative challenges of cooking, home brewing, and photography.
Be the first to know!
Never miss a story! Get the Outsmart newsletter.