The scale behind that finding holds steady from year to year. In its June 2025 assessment, GAO reported that combined cost estimates rose $49.3 billion, or 8.3 percent, across 30 major programs it had also reviewed the year before. The portfolio it tracks runs to 106 programs and roughly $2.4 trillion and the average time for a major program to deliver even an initial capability now sits at almost 12 years from start.
Every program office knows this statistic and works against it. To compensate at the program level, leaders need to monitor cost and scheduling to recognize drift while there is still time to act. This process is called earned value management.
Comparing planned value, earned value and actual cost gives a program objective forecast data on whether spending is running ahead of the work completed and a cost performance index that begins sliding early in a program reliably predicts an overrun at the end. This method relies on the data that is input and tracked during the lifecycle of the program. Earned value reflects the baseline behind it, so a weak baseline yields confident figures that describe the wrong plan. It also reports on the accounting cadence, usually monthly. By the time a cost variance appears in a formal report, the hours behind it were worked several weeks earlier.
There is an earlier place to look and most programs already collect the data. Labor is the largest cost element in the work these programs perform and it is usually the base over which indirect costs are allocated. When labor distribution moves, it moves both the direct charge to a contract and the overhead and general and administrative rates that ride on the labor base, so a change that looks small on one timesheet shifts the cost of the contract and the rates applied across the business. Labor data carries more signals per hour recorded than almost anything else a program tracks.
Labor data is captured every day, because compliance requires it. Government contractors record time daily, charge it to specific contracts and account for all hours worked. The records that satisfy an auditor also describe, in close to real time, what the workforce is doing. Read as a stream rather than a monthly total, three patterns tend to surface before a formal variance does.
1. Charge-code drift. Work migrates off the codes it was planned against and onto others, often for practical reasons that no one flags at the time. The plan and the effort separate and the separation is visible in the charging data weeks before it registers as a schedule problem.
2. Overtime concentration. A single team starts logging longer weeks to hold a date and in a monthly aggregate that can sit inside a total that still looks acceptable. In the daily stream it shows as a specific group pulling ahead of its plan on effort, which is among the clearest early signs that a milestone is at risk.
3. Labor mix. Senior hours substitute for junior hours, or the reverse and the blended cost of the work changes even when the total hours look right. Headcount and total hours can hold steady while the composition underneath them moves, so an hour-count based view misses the shift. A program watching the composition of the hours sees the cost of the work changing before the invoice confirms it.
-
Read as a stream rather than a monthly total, the same record becomes a leading indicator of program health.
Uncompensated overtime is another example case of a labor pattern that carries both a cost consequence and a program signal. When salaried, exempt employees work more than a standard week and record only forty hours, the recorded distribution overstates the cost of whatever project absorbed those forty hours and understates the effort that went in. Recording all hours worked, paid or not, is what total time accounting requires and its absence is among the most common findings in a labor audit, because it distorts the labor base that direct charges and indirect rates are built on. An engineer working fifty-five hours and booking forty is a sign of a team under strain and the gap between hours worked and hours recorded measures that strain directly, well ahead of a slipped date.
This does not replace earned value management, which remains the discipline that ties labor and cost to a measured plan and produces the formal record a program is accountable for. Reading labor data continuously sits ahead of that record. It gives a program leader a few weeks of warning to ask questions while the answers can still change the outcome and it makes the eventual earned value report less likely to contain a surprise.
Performing earned value management and labor tracking does not call for new data collection. The information is already gathered, validated and stored to meet compliance obligations. Most programs treat the labor record as something produced for the auditor and the monthly close, reviewed after the period ends and then filed. Read instead as a continuous stream, the same record becomes a leading indicator of program health, available weeks ahead of the earned value report and at finer resolution than any monthly total.
The discipline this asks for is modest and mostly organizational. It means giving program controls access to labor data at the cadence it is recorded, agreeing on the few patterns worth watching and reviewing them before the accounting closes rather than after. The data already exists on nearly every program, gathered for other reasons. Reading it as a program signal is a matter of practice more than technology.
