Predictive analytics: What HR teams need to know

Picture this: It’s your job to keep an eye on employee retention at your company. Thanks to workforce data and artificial intelligence, you have tools that can help flag who might be a flight risk. And these tools give you a heads-up; for example, you could be at risk of losing a group of high-performing women who are poised to take on leadership roles. 

Thanks to predictive analytics, you have an opportunity to take steps to retain those high performers. Predictive analytics can also alert you if, for example, you don’t have enough employees with the right skills for next year’s big initiative. And it can be used to promote fairness: Before you make a big offer to a job candidate, analytics can tell you if it will throw off your payscale, perhaps by making employees of colour underpaid in comparison.  

“You can ask, ‘Are we building glass ceilings in our company?’ and ‘Is there a ceiling where we just don’t see the same level of representation that we see in other levels?’ And you can identify that with predictive analytics,” says Trey Causey, head of AI ethics and senior director of data science at Indeed. “In fact, I would be hard-pressed to think about how we would identify that happening without using these tools. This is a way to be able to approach these questions quantitatively.” 

The proliferation of general analytics tools isn’t new to HR; they started gaining ground sometime in the second half of the last century. Today, the majority of large companies rely on analytics to look at data on past performance and inform decisions around hiring, firing and promotions. 

However, the rise of big data and artificial intelligence (AI) means that HR no longer has to rely on lagging indicators. Predictive analytics, coupled with new and innovative AI-powered technologies, can look ahead. How would it impact people if you make every Wednesday a meeting-free day? When should you consider offering an employee a bonus to keep them engaged? 

Although many of the early adopters of novel tools are big tech or financial services companies, uptake is increasing across industries, according to Dr. Hallie Bregman of the Bregman Group, an HR consultancy focused on data-driven HR practices. “This is absolutely the beginning, and it’s going to grow,” she says. “Five years from now, we’re going to see a lot more of it. Ten years from now, everyone will be doing it.”

To use these tools to their potential, HR teams should strive to be proactive, informed and responsible with data, so that the technology helps employees instead of undermining them. 

Follow these six best practices to make sure your company is implementing predictive analytics the right way.

Know what you’re solving for

“I can’t tell you how many clients I work with who don’t know what question they’re trying to answer” through predictive analytics, Bregman says. “They just say, ‘I’m supposed to be doing it,’ but they don’t know why.” 

“The big thing to ask is, what problem are you trying to solve in the next three to five years?” she says. Is it building out a new product team? Increasing diversity? Fuelling major growth? If you don’t clearly identify what you want to achieve with predictive analytics, you can spend time and money on tools that aren’t helping you hit company goals. 

Talk to your legal team

It’s also important to consult your legal team. They can evaluate whether the tools you’re using are sharing data in line with company rules, as well as privacy and anti-discrimination laws. “I’ve walked into far too many organisations only to see raw data being disseminated to groups of people that don’t need all the attributes on a report, which starts to bring risk,” Bregman says. 

Lean into the science

Not all predictive analytics products will align with your company’s values and business ethics. It’s essential to assess employees based on bona fide job requirements and data related to their behaviour and performance. Don’t base decisions on personality assessments that predict employee loyalty by considering whether someone identifies as a cat or dog person, or is from a small town. 

“There are a lot of vendors selling pseudoscience dressed up as predictive analytics,” Causey says. “Don’t ever be afraid to call in a second opinion, somebody who has expertise in the area.” If your people operations team doesn’t have that capability, look for services and consultants that evaluate HR tools in an unbiased way.

Keep in mind that you want tools that evaluate large datasets for trends. Try to avoid using granular data, such as one employee’s performance in a single quarter – particularly when there may be underlying circumstances the analytics tool knows nothing about. 

“It’s almost like the stock market – you don’t want to day trade,” Causey says. “You don’t want to overreact to small blips in metrics. You want to look at what the long-term trend is and make sure that you’re also using context as part of your decisions.”

Take the initiative on DEIB

Predictive analytics offers the opportunity for data-driven evaluation of practices on diversity, equity, inclusion and belonging (DEIB), which can otherwise be hard to quantify. With responsible privacy protection in place, you can take a data-based approach to help analyse decisions on salaries and raises, training opportunities and promotions. The goal is to ensure that compensation and development opportunities are aligned with performance and skills – and not driven by a hiring professional’s gut instincts, which can be affected by unconscious bias. 

Analysing employee data can also help HR teams ensure that opportunities are offered consistently across demographic groups. “If they’re not, what’s going on?” Bregman says. “Can you intercept that?” 

This type of data can make it easier to make a compelling case for DEIB at a time when some business leaders are discounting or dialling back such initiatives. “If you can use the data to demonstrate that X, Y or Z is happening, rather than just arguing from principle, you’re much more likely to be a change agent than if you’re making an impassioned plea because it’s the right thing to do,” Causey says. 

Check for bias

“Remember that the data we use to train algorithms and AI tools can be biased data,” says Dr. Salvatore Falletta, a professor and program director for human resource development at Drexel University. “And then the AI tool is as biased as a human – sometimes more biased, because it’s amplified.” 

If you have in-house tools, make sure bias checks are performed. If you’re buying from an outside vendor, ask how they have evaluated their products for bias. 

Don’t forget that data is just one input in the decision-making process, Causey says. If you don’t agree with the machine’s output, you don’t have to follow it unquestioningly. If predictive analytics sets recruiting targets too high or recommends timelines that are too short, adjust them.

Don’t be creepy

When leveraged in a benevolent way, predictive analytics can level playing fields, implement better hiring practices and make employees feel valued. But taken too far and without regard for privacy, predictive analytics can be downright creepy. What if such tools are monitoring social media updates as data inputs for whether an employee is a flight risk? Or tracking people’s movement around the office to see who they talk to in a day and how communicative they are? Or putting 'bossware' on remote workers’ computers to score their productivity? 

Causey, Falletta and Bregman all stress that it’s imperative to be transparent with employees about what data you’re gathering and why. No one wants to feel like they’re being spied on. 

“Think of it like the golden rule: How would you feel if you were being evaluated this way?” Causey says. “We certainly don’t want to evaluate employees in ways we wouldn’t like to be evaluated.”