Five or six years ago, companies realized that they were sitting on a wealth of data about their own employees. "People started to realize, 'Wait a minute, there’s a lot of data in here that we’re not using. Some of it is wrong. It’s not very clean,'" says Josh Bersin, the founder of Bersin by Deloitte, an HR research and advisory arm of Deloitte. "But if we look at it like we look at customer data, we could probably make much better decisions about who to promote, where they should be in the company, what role they would be successful at.'"

Since then, the people-analytics industry has emerged, with companies using algorithms and Big Data to recruit and assess employees. One report from McKinsey Global Institute estimated that social technologies, such as internal networking tools, can boost not only employee happiness, but also productivity by up to 25 percent.

By now, there are a handful of startups armed with software and surveys to measure a wide range of employee statistics. For analytics on employee engagement, there's Culture Amp, BlackbookHR, and RoundPegg. Companies such as BetterWorks game-ify goals at work. Retailers and banks have developed models to predict which employees are on their way out of the company. One company, Sociometric, measures physical proximity between team members and crunches productivity numbers.

But as the industry grows, big questions remain about what can be done with this newly discovered trove of data. Bersin's research shows that only four percent of large companies can make meaningful predictions about their workforces, while 90 percent can accurately predict business metrics such as budgets, financial results, and expenses. Can human-resources analytics do enough to capture the behavior and preferences of its endlessly complex subjects: humans?

"It’s one of the few areas of business that hasn’t really been figured out yet," says Bersin. "People are imperfect machines. Nobody ever figures out people completely."

But that doesn't mean companies aren't going to try. On the Big Data front, the company VoloMetrix mines calendar and mailbox data to determine over a hundred predictive indicators. From those indicators, the company works with clients to determine how to solve a given problem, from determining what makes a great salesperson to how emails can be more efficient.

"There are several different types of clients who work with VoloMetrix," says Ryan Fuller, the company's CEO and co-founder. Fuller says that VoloMetrix's clients either have a specific issue they want to employ data mining to solve, or hire the company to look more generally at how to save time. "Once the people analytics data is available, firms can immediately begin making data-driven decisions to improve efficiency and performance," says Fuller.

Some of the surprising results VoloMetrix has found from client datasets challenge conventional workplace wisdom. For example, for a client that wanted to know when the best time of the day was to have meetings, VoloMetrix looked at how disengaged employees were by seeing how many emails they were sending during meetings. At 9 a.m. meetings, roughly 8,500 emails were sent, while meetings at 6 p.m. were only slightly better at 7,000 emails. Meanwhile, employees in meetings between 10 a.m. to 2 p.m. didn't send very many emails—so the company rescheduled for the middle of the day.

In another study, VoloMetrix found that the best employees tend to not only have larger networks within companies than other employees—they also engage more with staff both more senior and junior than themselves. "High performers engage in more of both activities," says Fuller. "Within some companies, high-performing employees engage with senior staff members 28 percent more than low-performers and directly engage with junior staff 16 percent more than low-performers."

As for the problems that Big Data can't solve, small data might help. The company TINYpulse works with 500 companies to take feedback surveys, typically a yearly chore, and turn them into a weekly, anonymous, one-question pop-up. Some of the questions they've asked have garnered some very unconventional, but perhaps incredibly honest, answers. For example: the question “If you were promoted to be your boss's manager in the new year, what's the first thing you would change?"​ The most popular answers ranged from traditional answers such as better pay and hours, to firing and demoting employees who were dead weight. Another unconventional question TINYpulse asks to measure workplace satisfaction is whether employees have interviewed for another job in the past three months.

"Traditionally HR has been about compliance and what you can’t do. And now I feel HR is about inspiration and what you can do," says TINYpulse's founder, David Niu. "It’s got to be simple...employers have to show they’re doing something with the data and not wasting people’s time."

In a TINYpulse survey of over 30,000 employees, the company found that 34 percent of the happiest employees say that they work with great people. Those sentiments are amplified by employers figuring out problems big or small—from compensation and layoff anxieties to the condition of the bathrooms or break rooms—by sending out quick, easy surveys: TINY's surveys have a 90 percent response rate at some of their client companies.

Yet even as new companies crop up to solve corporate problems with numbers and analytics, the HR department won't disappear just yet. The recent criticisms with big data has been that in the absence of theory, it doesn't do much good. In other words, numbers don't speak for themselves. Bersin says that while there's a big maturity gap between companies that are incorporating people analytics into their HR departments, both venture capitalists and company managers are keeping a close eye on the analytics companies that are innovating. As data and information becomes abundant, the mystery of how a particular employee fits into a company, or what solutions a particular company needs will still require a human touch.

"It’s very tricky. People data can be very misleading," says Bersin. "The data won’t necessarily tell you everything: You have to interpret it, know what it means, and try to make sense of it. It’s not like you can sit in a black box and look at the data."