The Navy’s commitment to an “innovation” culture is mostly hollow rhetoric. Seemingly endless talk using words such as “rapid” and “innovative” serves bureaucratic goals such as securing funding for programs, but truly innovative outcomes are rare and almost never rapid. The Navy is in a state of self-delusion, believing it is innovating at all levels when in reality it is often standing still. There is a massive thinking-doing gap.
The Navy has countless innovation “champions” who are actually risk averse. They appropriate the latest innovation buzzwords while remaining comfortably ensconced in the laborious requirements-building system where they can avoid risk and wrap themselves in bureaucratic red tape rather than prototype and test their ideas against viable alternatives.
The Navy must not accept this. Innovation means nothing until working prototypes compete against one another with the potential to scale. To truly change, the Navy must create ideas-to-action processes that allow ideas from all corners to battle for backing and implementation.
Following this line of thinking fosters a bias for action. In the spirit of putting our money where our mouth was, we set out to create a product for the Navy and found some traction with one key leader.
Getting to Work
In an article for War on the Rocks, our small prototype team, Naval Algorithms, argued that the Navy did a poor job of allocating talented sailors to billets where they were most needed. We offered a roadmap to redesign the process using algorithms as the building blocks for the Department of Defense’s (DoD’s) dream state of operations powered by machine learning and artificial intelligence. The Chief of Naval Personnel (CNP), Vice Admiral Robert Burke, read the article and noted that it aligned with his existing initiatives. He invited us to Arlington, Virginia to learn more about the Navy’s efforts to transform the Manpower, Personnel, Training & Education (MPTE) enterprise, especially Sailor 2025 and its Detailing Marketplace initiative. After meeting with a variety of stakeholders, we presented him with a roadmap to develop and prototype an algorithm-based manning model.
He immediately greenlighted the project to quickly create and pilot a new detailing system. Admiral Burke sent us on temporary assignment to the Defense Digital Service, the Pentagon’s software modernization arm. In five weeks and for a cost of less than $5,000, we delivered an algorithmic system that built on and further advanced previous Army and Air Force efforts to mathematically optimize matching servicemembers to jobs. Other services have spent millions of dollars on the same effort.
We tested the mixed-integer programming (MIP) algorithm, using preference data from the medical, cryptologic warfare, and explosive ordnance disposal communities. Put simply, we took the preferences of sailors and matched them to the preferences of commanding officers. Nearly everyone got more of what they wanted from each side of the equation, and that was possible because of computers running thousands of iterations of thousands of different possible realities in ways that humans merely cannot do quickly.
The team’s MIP algorithm outperforms human detailers and previous algorithms, especially in later preference matches. MIP creates every potential lattice the system could ever create, then selects the best one. Previous algorithms (deferred acceptance) try to build the most optimal lattice in one go. Humans can do well at first, but quickly fall off in optimization due to the iterative complexity involved (see the graph for how much better the MIP algorithm outperformed the human detailing standard for the three Navy communities).
The algorithm was specifically designed to be integrated back into Admiral Burke’s existing vision for a Detailing Marketplace. Looking further, our backend algorithm ought to be combined with a front-end. We suggest linking with other Navy efforts like Jetstream, the explosive ordnance disposal community’s frontend detailing marketplace effort supported by the Defense Innovation Unit (DIU).
Turning invention (our algorithm) into innovation (adoption and integration with Navy communities) requires both quantitative and qualitative efforts. Moreover, we believe that other lean teams with high-level support can repeatedly turn ideas into innovative processes if given the time, tools, and top-cover.
Building a viable prototype while working a traditional day job is nearly impossible. The team must be 100% focused so it can complete the objectives in a limited amount of time. However, autonomy does not mean a lack of accountability. Our team had a liaison at the supporting office (in this case, the Office of the Chief of Naval Personnel) who served as a bridge to the innovation team (us working out of the Defense Digital Service), while still allowing us to define our own milestones, move at our own speed, and share our progress.
Lean teams cannot be hamstrung with poor tools. Because talent management is a challenge for all the military services, working from the Pentagon allowed our team to liaise with Air Force and Army teams and access their data and learned lessons. Within the defense apparatus, our project required coding that would be impossible on Navy systems, so we worked from the unique Defense Digital Service spaces and used computationally fast computers (such as Macbooks) and commercial-grade connectivity speeds (such as unrestricted Wifi), both extremely rare in the Pentagon or any other DoD space.
No commander will love temporarily losing members of their unit, so a higher-ranking officer must take the risk to pull sailors to work on projects that can make a grander impact beyond their day job. The lean team must be able to work on the project uninterrupted by outside requirements. In our case, each team member’s commanding officer quickly approved the temporary assignment.
The output of a lean team’s innovation sprint should be a product that can be tested and evaluated with the goal of integration into existing systems, or a well-conceived reason for starting fresh. Our manning algorithm was not intended to replace the current sailor detailing system, but rather to improve it. With this in mind, we read key Navy manpower documents such as the “U.S. Navy Health of the Force Report, Growing to Win: Sailor 2025” and the 2019 “Manpower, Personnel, Training & Education Strategic Goals” to align our work with the longer-term effort of a detailing marketplace.
Senior leaders often recognize the need for ideas-to-action pipelines, but they still need a standard playbook to follow. This talent management pilot project—enabled by time, tools, and Vice Admiral Burke’s top-cover—shows a repeatable path for turning ideas into durable innovation.
All authors graduated with distinction from the U.S. Naval Academy and together make up Naval Algorithms, the small prototype team described in the article.