Weapons-maker Alliant Techsystems Inc. has developed a workforce planning system that’s like a rifle scope with night vision; look through it and problems can be spotted in surprising places.
Last year, for example, human resources officials at the company’s armaments unit examined a range of factors using statistical analysis to study the “flight risk” of each of the roughly 1,200 employees in the group’s Radford, Virginia, operation. Organizations often figure the main threat to future success comes from departing executives. But one of the biggest challenges facing the Radford unit was the potential defection of an employee below the management level. The worker, an engineer, was central to a major initiative of the armaments unit, and his age and tenure were both flight-risk factors.
“He wasn’t a manager, but he was critically important,” says Cory Edmonds, manager of human capital analytics at Alliant Techsystems, which is better known as ATK. He popped up as a “high risk.” The company then began to prepare for the employee’s departure, considering, for example, assigning a junior engineer to work as an apprentice to him. Ultimately, the engineer didn’t leave and ATK didn’t take any actions. However, if the employee appears as a flight risk again the next time the company runs the numbers, HR officials will encourage business unit leaders to take steps to prepare for the worker’s departure.
In recent years, such sophisticated data analysis has become a workforce defense weapon at ATK. The Minneapolis-based company has become more adept at spotting and overcoming talent management challenges. In doing so, the 18,000-employee firm has put itself at the forefront of the new frontier of metrics, as planning tools increasingly allow companies to predict workforce needs.
Such prophesies promise to be profitable. Carl Willis, vice president of human resources at ATK, expects workforce analytics to save the company hundreds of thousands of dollars over several years in reduced hiring and severance costs and in greater use of contingent and temporary labor during peak production cycles.
To achieve such gains, ATK has tapped outside expertise, including workforce analytics specialist Orca Eyes Inc. The company also has fortified its HR department with quantitative experts. Edmonds, for example, majored in statistics at Virginia Tech and has taken graduate courses in the field.
“If you really want to do this kind of work, the typical skill set you have in HR is not what you need,” Willis says. “You need more of an analyst.”
What started as an experiment in one division is now a companywide priority with the full backing of ATK president and CEO Mark DeYoung. The decision to expand the workforce analytics effort throughout the company was made before DeYoung took the reins last year, but he calls the analytics initiative “very important.”
Not all of DeYoung’s executive team members employ the flight-risk model.
Nonetheless, he expects those predictions to be used more widely as more results and success stories are shared. “The aerospace industry is well-known for its aging workforce and we need to look beyond the horizon to ensure that we have the right talent to lead our company long after I’m retired,” DeYoung says. “Analyzing the needs of our business, developing the talent to meet those needs and identifying recruitment candidates is the job of every senior manager.”
There have been bumps along the way, including questions about how to handle sensitive findings, but the company is charging ahead with its data detection system. That puts ATK in the minority. Only a small percentage of firms have significant capabilities to predict the future of their workforce and plan ahead. But experts expect more companies to follow ATK’s lead. Predictive analytics efforts fuel bottom-line success, says Dan Hilbert, CEO of Orca Eyes, which has a partnership with Workforce Management related to workforce data analysis. “You’re creating massive competitive advantages.”
ATK first tapped Hilbert’s company in 2008, when Willis was determined to improve the recruiting function of the armament systems group, one of ATK’s four main units. A 30-year veteran in HR, Willis has a background in Six Sigma methods of continuous improvement and liked what he had read about Hilbert’s work as an HR official at oil company Valero Energy Corp. There, Hilbert had fashioned recruiting into something like a manufacturing supply chain complete with alerts about problems with hiring quality or speed.
In 2007, Hilbert founded Orca Eyes with a vision of making workforce data analysis more widespread through software and consulting. When Willis came calling, Hilbert found in ATK a company very receptive to workforce analytics. “This is an organization that would have done it without us,” he says.
Orca Eyes isn’t the only vendor pitching software and services for HR data analysis. Competitors include technology heavyweights IBM Corp., Oracle Corp. and SAP, along with smaller players such as SuccessFactors Inc.
Workforce analysis has been demystified in recent years by books such as Moneyball, which showed how the Oakland A’s achieved success on a shoestring budget by focusing on lesser-known statistics such as a hitter’s on-base percentage.
Thanks in part to such accounts, workforce analysis has emerged as a priority for many organizations. Studying data around recruiting, performance and turnover can lead to insights such as the best sources of new candidates, the most effective training programs and the most valuable employees to retain. Through regression analyses, organizations can determine the relative importance of different factors in an outcome. These relationships are all but impossible to detect at first glance but often make intuitive sense in retrospect. They also allow companies to take the next step in workforce metrics: making predictions about the future.
At ATK, Hilbert helped persuade Willis to hire a workforce planner as well as the statistician Edmonds. Together, the team created a system for predicting job openings by looking at the future supply and demand of talent for the armament systems group, which employs about 4,600 people including the Radford facility. For the demand piece, the team created head count projections based on company expansion targets, and then checked these with business-unit leaders.
The supply side of the equation is where the heavy-duty data-mining came into play. Traditional methods of predicting turnover apply past attrition rates to future scenarios, and they typically don’t delve into details beyond the level of business units or broad job categories. ATK and Orca Eyes attempted to take planning to a new level. In particular, Edmonds used regressions to create statistical probabilities for attrition for each individual employee. The team then “rolled up” those individual flight probabilities into risk assessments.
A granular view is important in part because ATK has so many specialty jobs. A large retailer such as Starbucks Corp. may have a majority of employees in a few roles, but 60 percent of the positions in the armament group have one person in a particular role at a specific location, Edmonds says. For example, a single worker may have expertise with small- or large-caliber ammunition or export-control rules.
Determining the stability of teams also is important to ATK. Those groups may play critical roles, such as the roughly 150 people who handle maintenance at the Radford facility. The ATK-Orca Eyes system succeeded in forecasting unusual turnover in that maintenance unit. In the 12-month period ended Aug. 31, 27 craft workers left the unit.
Historically, only about 10 people had left the unit annually, but the ATK-Orca Eyes flight-risk model predicted 21 would go. Key to the prediction was the phase out of a health care benefit for retirees in 2010, which gave veteran employees an incentive to quit. Edmonds had studied the effect on attrition of another retirement benefit that expired some years ago and had factored that into last year’s projections. Traditional forecasting systems, he says, would have focused on prior year turnover rates of the unit and perhaps a couple of other factors such as job category or age. As a result, Edmonds says, conventional methods “would have missed the retirement medical benefit change.”
Another example of ATK’s prediction prowess involves production workers at Radford who perform such duties as mixing chemicals and cleaning scrap metal. Last year, ATK’s flight-risk system correctly projected an unusually low rate of turnover among production workers. Traditionally, nearly 100 would leave in a given year. But the ATK-Orca Eyes software forecast that just 61 would leave. In the end, 65 departed.
“The reason it was so much lower,” Edmonds says, “is because we had relatively few hires over the year before compared to previous years.” New production hires tend to leave at a relatively high rate, while those who make it past their first year are more likely to stay. The ATK-Orca Eyes system accounted for this by considering the tenure of each production worker.
The successful predictions of unusually high turnover among maintenance workers and unusually low turnover among production workers did not translate into concrete actions last year. Those were among the first forecasts of the flight-risk system, and “we did not have as much confidence in the model,” Edmonds says. Now that the system has proven valid, ATK plans to use future results to prepare early for unusually large amounts of hiring and other workforce demands.
Edmonds’ training lends itself perfectly to this sort of analysis. As a graduate student, he worked on a project to predict traffic accidents by sifting through a range of factors in past collisions, such as vehicle speed and whether drivers were keeping their eyes on the road. Now at ATK, Edmonds and others on the team have similarly tried to isolate the relative importance of various factors on turnover. Using regression analysis, they have plugged in usual suspects such as employee age and tenure on the job. They also have explored less conventional factors, such as the local unemployment rate, marital status and time of year workers left ATK.
Initially, Edmonds had retrieved data on the day that employees left the company from ATK’s HR information system in order to study the attrition effect of the local unemployment rate, which changed over time. Once the team had that data, it also decided to see whether people are more apt to leave at certain times of the year for other reasons. There was indeed a trend: ATK employees tended to retire in the second and third quarters of the year. “Most people want to retire before the winter hits,” Edmonds says. “That’s how we explain it.”
The time-of-year insights weren’t the only surprises. The team also found that single people are much more likely to leave through involuntary attrition than married employees. Yet, such discoveries raise thorny issues. What is ATK to do with information about marital status? It can’t very well start giving preferential hiring treatment to married people.
ATK says it doesn’t use its data-mining findings to discriminate. Instead, it taps the data to lessen flight-risk dangers, such as by setting up mentoring relationships and trying to improve employee engagement. “This information does not go into any decision around termination or promotion or in our external hiring, but it does help the company mitigate risk proactively,” the company said in a written statement.
Edmonds notes that flight risk is a complicated calculation. “Just because someone is unmarried doesn’t mean they have high flight risk,” he says. In addition, ATK says it keeps flight-risk data highly confidential.
Still, determining a high flight risk for an individual contributor begs a question: Should the company directly ask workers about their retirement plans? ATK doesn’t take this approach. “The point is not to confront employees and influence a decision that is theirs,” Edmonds says. At the same time, he concedes ATK remains in the early stages of understanding how to respond to some workforce analysis results.
“A lot of people struggle with strategy, and translating strategy into tactics,” Willis, the HR vice president, says. “This is a process that can translate strategy to action and the bottom line. And that’s what’s so cool about it.”
Workforce Management, March 2011, pgs. 22-23, 26 -- Subscribe Now!