Many top industry names like Google, Cisco and Sprint are starting to use predictive analytics to aid in critical workforce management functions, like hiring, retention and salary decisions. So why not the Navy? Every day the Navy must make the decision of either training a pre-existing employee for a needed task or hiring someone new for the position.
For two Naval Postgraduate School (NPS) researchers, Dr. Amilcar Menichini and Dr. Thomas Sae Young Ahn, who teach in the university’s management programs and support the Acquisition Research Program (ARP), thinking about analytics and these types of decisions began a few years back. As the ARP expanded over the years as a platform for improving analytical effectiveness and problem-solving for Department of Defense (DOD) acquisition strategies, it helped inspire a five-year Acquisition Workforce Strategic Plan initiated in 2016 to help address retention of mid-career employees in the DOD.
Menichini and Ahn think predictive analytics can help maintain a stable workforce. The two formed an interdisciplinary partnership, with Menichini bringing experience in finance and Ahn in econometrics. With funding by the Naval Research Program (NRP), the pair is now in the middle of creating the Dynamic Retention Model (DRM), a predictive analytics model designed to create a hypothetical office full of individuals with different motivations, skills, experience, etc., that is then introduced to different scenarios, such as a pandemic or high turnover. Not only would a Navy command be able to look at the workforce quality as a whole, but they could also narrow down to look at how each employee might respond to different scenarios.
“[The employer] can come up with their own little experiments and the simulation will run it and give them the tools they can use to set personnel policy in the future,” Ahn explains. “It’s going to allow them to, in some sense, look forward and sort of wargame it out by introducing shocks and changes to pay structure.”
In the case of a pandemic, for example, the employer could plug into the system a spike in the civilian sector unemployment rate. Then the program would probably determine the employees will stick around in the government job because it’s not a good time to look for work. But when the pandemic nears to an end, the employer might be more at risk of losing some employees. Then the question is, would a bonus or pay raise do better at ensuring retention? How much for how long? These answers can change depending on the employee’s length of employment and experience level, among many other factors.
The Dynamic Retenion Model would look at all the options based on different speeds of economic recovery to determine the likelihood of retaining employees. Not only will employers be able to monitor the retention rate of individuals, but personnel policy leaders would be able to determine the quality of the workforce as a whole, and how diverse it is in age, race and sex.
“If you’ve got a menu of items that you want to maintain and grow about your organization, then using a model like ours will not just drill down to a particular aspect, but actually look at how the whole workforce moves and evolves,” Ahn says.
Menichini and Ahn hope Navy employers could come to DRM when making any recruiting, promoting or personnel decisions, and personnel policy leaders could use it before adjusting policy to better see the long-term impacts of these large decisions. The code won’t make the decisions, but it willll help inform decision-makers.
“It’s really about providing simulations and best guesses for the future so that decision-makers can have a full quiver of arrows and aren’t just shooting in the dark,” Ahn explains.
The researchers see this tool as a way for the Navy, and DOD as a whole, to analyze every acquisition and retention decision holistically, which is especially important for the government to do because it’s slower to change than a private entity. They think it will help different departments proactively adjust rather than just react.
The NPS pair is now in the process of coming up with all the variables for the model to play with and coding in data (using MATLAB). The more data from past scenarios the program has to work with, the better the model can work.
“We’re proud of this research and it has academic value, but it would be a tragedy if it just ends up in a journal somewhere,” Ahn stresses. “We envision this research agenda as not retrospective, but prescriptive. We did this research because we thought this can really contribute, in our own small way, to national security.”