This article originally appeared at RealClearDefense on April 29, 2016. It was co-authored with Alexander Kirss, a resident junior fellow at the Center for the National Interest.

Congress’ recent interest in reforming the Pentagon’s acquisition system is laudable, but its efforts will be wasted unless it addresses the root-cause of the system’s dysfunction: the requirements definition process. Requirements definition is the aspect of acquisition where the technological capabilities a new weapon will have are determined. Problems in the requirements phase of the acquisition process are therefore carried forward and contribute to a host of far-ranging issues down the line.

Those issues later in the acquisition process are evident in a recent Government Accountability Office (GAO) report assessing selected acquisition programs. GAO found sixteen cases where acquisition programs saw major increases in development costs during the production phase of the process. Since this phase occurs long after the capability requirements of a program have been set, development costs should stay relatively flat. But in eleven of sixteen cases, increases stemmed from the addition of unplanned capabilities—while the other five occurred due to deficiencies discovered in testing. Upfront cost increases for capability upgrades totaled $3 billion, but the new requirements add complexity to the production process that will balloon costs later.

There are three possible explanations for these capability revisions. First, changing operational requirements could mean that the originally planned capabilities no longer meet current operational needs. Second, new or emerging technologies could have become available following the initial requirements definition phase. Third, bureaucratic log-rolling by the military services and their subcomponents allowed for unrealistic requirements at the outset of the process that had to be revised later.

While the first two explanations do not stem entirely from the requirements definition process, they are affected by it. The fact that capabilities determined during the initial planning process might need to be changed to account for changing operational needs suggests that both the requirements and production processes are moving too slowly. The same goes for the introduction of new technology late in the process.

The third explanation ties into the first two. Bureaucratic log-rolling allows for overstated capability requirements that are really technological “wish lists,” because no service has an incentive to veto another service’s plans, no matter how costly or unfeasible they may be. The military services want weapons that push the technological envelope, and attempt to predict what will eventually become state-of-the-art technology five to ten years out with little actual knowledge of whether their dream can become reality. Because there is little oversight in the Joint Concept Development process and Joint Capability Integration and Development System, the services and their subcomponents have tremendous leeway in pursuing their own priorities. The Joint Requirements Oversight Council (JROC) should provide a check against excesses but rarely does. James Locher, a retired military officer and former influential Senate Armed Services Committee staffer, previously referred to the JROC as a “rubber stamp.”

Absent large-scale changes in Title 10 of the U.S. Code, it is unclear how to break up the service cartel. As it stands now, each of the services has equal representation and effective veto authority on each of the major requirements oversight committees. When controversy does arrive, back-channel communications at the flag officer level usually settles it.

Ideally, civilians in the Office of the Secretary of Defense (OSD) should have authority to provide oversight and demand the services prioritize. But even then civilian oversight would be difficult if the services can present a united front. Despite his many sins as secretary of defense, Robert McNamara played the services against each other to good effect. Since then, the services have banded together under the banner of jointness to prevent a similar situation. Therefore, addressing requirements means both increasing OSD oversight authority and addressing the incentives for the services to not only log-roll, but also to present a unified front that stymies civilian supervision.

Congressman Mac Thornberry’s latest acquisition reform proposal, which mandates that the services chiefs must certify that requirements are “realistic and necessary” before submission to the JROC, does not ignore requirements completely.  But if a service chief is pursuing the service’s organizational interest, and the organizational interest is to continue to pursue requirements that push the technological envelope, and service log-rolling facilitates the pursuit of those requirements, then it is unclear how this actually addresses the problem.

Legislators are not the only ones to overlook this problem. In recent expert testimony on defense reform before the Senate Armed Services Committee, fifteen witnesses offered twenty-two separate recommendations to fix the acquisition process. Only one expert witness addressed requirements.

Those interested in fixing the Pentagon’s acquisition system need to actually address it. Requirements reform serves the goals of current acquisition reform effort and ignoring it will it inhibit the effectiveness of current proposals. It will not be easy, but it should take greater priority than addressing the other phases of the acquisition process. The requirements process feeds into and ultimately drives the other phases of acquisition. Addressing it will target a root cause of acquisition dysfunction rather than just treating a symptom.

Matthew Fay is a defense policy analyst at the Niskanen Center and a political science PhD student at George Mason University’s School for Policy, Government, and International Affairs. Alexander Kirss is a resident junior fellow at the Center for the National Interest, he earned his MA in International Relations from the University of Chicago.