What Meteorologists and HR Professionals Have in Common

Evolv attritionChances are you haven’t thought of the morning weather report as an example of big data at work. But it is.

Pulling together information from ground weather stations, satellites, ocean buoys, and historical records, meteorologists are able to offer a remarkably accurate prediction of what today’s and tomorrow’s weather will be. Even the seven-day forecast has become far more accurate than not, as the amount of data that figures into each prediction has grown over the years.

This is the growing field of predictive analytics, which, in the case of weather, influences the price of energy on Wall Street, and the supplies a utility lays in, not to mention how you dress for the day ahead.

For years companies have used big (or sometimes little) data to make predictions about everything from the amount of inventory to buy, to the most efficient routing for deliveries, to sales projections for the next quarter and next year.

But it’s only been in the last few years that big data and the predictive value it holds has made its way into the human resources department. Credit Billy Beane and Moneyball for spreading the gospel of statistics. The Atlantic discussed this trend in detail a few months ago.

The potential is enormous. Consider the value of knowing whom among your most skilled workers was likely to leave in the next three months? But exactly what factors, what data, do you take into account to predict that or, for that matter, to predict anything at all?

The biggest companies have analytics groups within HR who make those decisions. They work with I/O psychologists and statisticians to find the patterns most likely to make accurate predictions and find hidden connections.

Now companies are emerging to provide that service. Evolv is one of the oldest in this very young field. This week it announced a major step forward in making its cloud-based predictive analytics tools, Evolv Insights, broadly available. In conjunction with Evolv Selection, managers and talent leaders can make far more informed judgments about the workforce than they ever could using even the best assessment testing available.

“We find patterns across the data,” explains Jack Mazzeo, Evolv’s senior director, product management. It’s a simple explanation of an enormously complex process that requires large amounts of data, regular feedback, and the development of  algorithms that are the secret sauce.

Just as weather prediction has improved as meteorologists and mathematicians discovered which of the millions of data bits exert more influence than others, so do Evolv’s algorithms. A company’s own data is a big part of the recipe, but the collected wisdom of other predictions is factored in as is publicly available data, such as from the U.S. Bureau of Labor Statistics.

“It’s a combination of data,” Mazzeo says. “We factor in a number of contributions … Evolv managementrun a massive number of permutations,” he adds, to “see a little bit into the future.”

Article Continues Below

All that occurs behind the scenes, of course. What HR professionals see are graphically presented results that, in the example Mazzeo and Carl Tsukahara, executive vice president, marketing and product, used, show what factors go into influencing a person’s decision to leave. And how much they contribute to the overall decision.

“If they want to understand the causal factors that are causing their attrition, we tell them,” says Tsukahara. If you want to know specifically who is at risk, Evolv tells you so, he adds, you can “act before the person leaves.”

In announcing the product availability, the company’s release observes that:

Evolv Insights allows managers to model different scenarios for achieving operational goals and then apply ‘what if’ analysis to forecast the impact specific changes will have on achieving them. As a result, companies can proactively improve the quality, productivity and performance of their workforce to align with business goals.

At the other end of the lifecycle, Evolv Selection works with assessment tools and and the collective company data to offer predictions about whom will make the best hire.

Alas, these tools and the services don’t come cheap or easily. They are not for small companies, whose data is too small, too incomplete, or too confused. For patterns to be discerned and be reliably used in making predictions, big data is needed. How big? Fortune 1500 size, says Mazzeo.

The cost to get started and set things up can run into six figures. Thereafter, the monthly expense is in the double-digits.

John Zappe is the editor of TLNT.com and a contributing editor of ERE.net. John was a newspaper reporter and editor until his geek gene lead him to launch his first website in 1994. He developed and managed online newspaper employment sites and sold advertising services to recruiters and employers. Before joining ERE Media in 2006, John was a senior consultant and analyst with Advanced Interactive Media and previously was Vice President of Digital Media for the Los Angeles Newspaper Group.

Besides writing for ERE, John consults with staffing firms and employment agencies, providing content and managing their social media programs. He also works with organizations and businesses to assist with audience development and marketing. In his spare time  he can be found hiking in the California mountains or competing in canine agility and obedience competitions.

You can contact him here.


4 Comments on “What Meteorologists and HR Professionals Have in Common

  1. Wow. Where to unpack this?

    “In conjunction with Evolv Selection, managers and talent leaders can make far more informed judgments about the workforce than they ever could using even the best assessment testing available”.

    Really? That’s a bold statement. Any actual evidence to back that up?

    “Just as weather prediction has improved as meteorologists and mathematicians discovered which of the millions of data bits exert more influence than others”

    That’s a pretty serious misunderstanding of forecasting. How the weather operates has very been well understood for a quite a long time, but the ability to model that understanding has grown with computing power- the “bits” so to speak are the physical areas the model is broken down from- what used to be 200 mile boxes is now 2000 foot boxes. This idea of discovering that some bits are more or less important? No relation to the reality of how it has evolved that I can see.

    Secret sauce? Perhaps this page may inform that concept:


    Finally, the Big Data faithful better get an understanding of this http://en.wikipedia.org/wiki/Emergence or they will surely find themselves on the rocks at some point…like Lehman Brothers et al.

  2. I would love to see big data be able to show CAUSAL results not just CORRELATION. We HR people have a big tendency to believe that just because something shows an impact on results means it CAUSES results. If this tool can do this I say HURRAH! But I am skeptical. Show me the proof of causality — that would be a tremendous breakthrough!!

  3. Jacque, at it’s best, Big Data can only provide a probabilistic forecast. Causality will never be absolute.

    The big problem with Big Data is applying it to emergent phenomena without understanding how that shifts the probabilistic curve, often beyond the range of usefulness.

    For instance, 10 day weather forecasts are still no better than straight probabilistic climate estimates, even though the models do run specific looking solutions. 3 day forecasts are much, much better than climate estimates.

    In human affairs, like the solution touted here, where are we right now on that scale? 9 days out, in my opinion.

    Also because of certain humanistic facts of life, hiring a weak performer is actually less of a risk than missing a top performer, but since you cant prove a negative and nobody wants to take flyers on hires, Big Data can actually decrease performance below random. Wrong information is much worse than no information. See again, the brothers Lehman….

    Powerful tools can be misused in powerful ways by the credulous, and that’s what it looks like from here…

  4. @ Everybody: good points, all.

    Hmmm. I wonder if the knowledge by its employees that a company had or could get such tools might reduce the tools’ effectiveness? Let’s say I were planning to leave my company inchoose, and I knew/suspected the company had something like this. I’d then work to avoid any “suspicious” activities that might flag my intentions…

    Lets consider another scenario: I am an important but (I think) underappreciated/underpaid employee. However, I like my company and want to stay- but just get the raise/promotion they’ve been stalling on for awhile. Since I know/suspect they use techniques as discussed, I do a lot of things to raise flags on my behavior, knowing/believing they’ll get me what I want. It’s like getting a counteroffer without even getting close to an offer!

    One way of having to guess when valuable employees might leave is to give them “multi-year, guaranteed-raise/bonus, no-layoff-without-cause” employment contracts. Funny how companies who say their employees are their major assets and quality hiring is their most important task don’t seem to take me up on that suggestion….


Leave a Comment

Your email address will not be published. Required fields are marked *