On Time and Under Budget: The Number the Government Does Not Publish

Question: What percentage of the contracts the current administration has entered has finished on time and under budget, and how does that percentage compare with the previous administration?

Answer: Nobody publishes that number, and no system exists that could produce it on request. The federal government tracks whether contracts exist, what they cost, and who holds them. It does not publish whether they finished on time, and it does not publish whether they finished under budget. The system that does record those facts is not public, is frequently late, and is the only place they live.

This is not a dodge. It is the finding, and it is worth stating plainly, because the question is being asked more often and answered more confidently than the record supports.

What a number like that would require

In fiscal year 2025, the federal government committed about $793 billion on contracts (GAO, 2026d). That is the denominator, and it is roughly stable from year to year. A defensible percentage would need three things: a record of every contract’s cost and schedule at close-out, a field in that record marking whether the contract beat its baseline, and a public tally.

Two of those three exist. Not in one place, and not for the same unit of account.

The first row is where most public argument draws its data, and it is the wrong instrument. USAspending can tell you that a contract was awarded, modified, and obligated. It cannot tell you that the thing the contract bought arrived late or cost more, because those questions were never fields in it.

The second row is the one that matters, and it is the one nobody can read.

The system that keeps the record, and the record of how it is kept

Federal acquisition regulation is unambiguous about where a contractor’s performance history lives. “CPARS is the official source for past performance information” (FAR 42.1501), and Subpart 42.15 requires an evaluation of every contract above the simplified acquisition threshold, prepared at least annually and at completion. The evaluation covers, among other things, the contractor’s record of “(2) Forecasting and controlling costs” and “(3) Adherence to schedules” (FAR 42.1501). Eight factors, two of which are exactly the question at hand.

Those evaluations are source-selection material. They are not published in aggregate, by law or by practice, and the Government Accountability Office has been reporting on the consequences for seventeen years. A 2009 report found that better performance information was needed to support award decisions; subsequent audits found the same gap. The Department of Transportation’s inspector general summarized the line in 2023: multiple GAO reports “have highlighted the lack of contractor performance information as an ongoing issue in the Federal Government” (DOT OIG, 2023).

What the Transportation audit measured in passing is the part worth pausing over. Of 120 sampled procurement actions, 89 were not registered within 30 days of award, and the department took “an average of 421 days to register contracts in CPARS, with less than 26 percent of the 120 procurement actions in our sample meeting the 30-day registration requirement” (DOT OIG, 2023). Some entries arrived more than two years late. This is one department, and it is not the worst of them. The 2017 Defense Department audit the same report cites found significant shortfalls of its own.

So the instrument that could answer the question is filled in, on average, more than a year after the work it describes, by officials with other priorities, at a compliance rate its own inspector general calls insufficient. Aggregate those evaluations and you would not be measuring contract performance. You would be measuring reporting performance.

What the closest published series actually counts

There is one place where a published federal series came near the question, and it is worth looking at both years of it, because the series changed what it measures.

The Department of Defense publishes annual assessments of its major information technology business programs, and GAO reviews them. In June 2025, GAO examined 24 programs representing $10.9 billion in planned spending. Of those 24, twelve reported cost increases, seven reported schedule delays, and, in the summary GAO published for the public, “10 systems were on time and budget” (GAO, 2025b). The longest delay ran to 48 months, with a median of 15 (GAO, 2025a).

One year later, the same series examined 18 programs representing $10.3 billion and stopped summarizing in that form. The 2026 report counts changes rather than successes: “officials for 11 of the 18 business programs reported experiencing cost or schedule changes since January 2024” (GAO, 2026c). Nine programs reported cost changes, five of them increases ranging from $2 million to $394 million and four of them decreases ranging from $247 million to $330 million; ten reported schedule changes, seven of them delays of 2 months to 12 months and three of them improvements of 17 months to 24 (GAO, 2026c). Three programs expect to rebaseline, meaning they will reset the standard against which they are measured.

Two cautions apply before anyone draws a trend from those two paragraphs. The cohorts differ, 24 programs one year and 18 the next, so the numbers are not a trend line. And note what “on time and budget” meant in the 2025 version: not that the systems were verified to have beaten their baselines, but that no cost or schedule change had been reported since January 2023. The absence of a change report is not the presence of performance. It is the closest thing to the requested statistic that the government publishes, and it is a count of unchanged programs in one department’s IT portfolio, not a share of federal contracts.

Why an administration-to-administration comparison cannot work anyway

Suppose the percentage existed. The comparison the question asks for still would not survive contact with how federal programs age.

First, the timeline. GAO’s 2026 weapon systems assessment, its twenty-fourth, found that “the overall average time frame to deliver a capability increased this year to over 12 years” (GAO, 2026a). A twelve-year average exceeds any administration’s tenure twice over. A weapon program signed in one presidency is delivered under the next two or three, and the delay visible today was scheduled into a baseline a decade ago.

Second, the selection effect. A contract signed this year cannot be late yet. The contracts whose overruns are visible today are, by definition, the ones old enough to have finished, and every one of them was entered years ago, under a different administration. Filter the universe by “entered by the current administration” and you filter out most of the population that could possibly have failed. The question selects for its own answer.

Third, the baseline moves. Programs that miss their targets can reset them. The 2026 IT assessment found three programs expecting to rebaseline and notes that repeated rebaselines may indicate a program is not managing expectations, or is facing unexpected complexity, or has trouble with its contractors (GAO, 2026c). Each rebaseline is a new chance to be on time.

Fourth, the pattern predates the question. GAO’s June 2026 roundtable report on acquisition waste states it at the scale of decades: the Defense Department’s “costliest weapon programs have consistently exceeded cost estimates and delivery schedules” since 1990 (GAO, 2026b). That span covers five presidents of both parties. The 2026 assessment found that between 2018 and 2025, 18 of 40 programs entered the fast-track middle tier of acquisition with immature technologies, a self-inflicted schedule risk that belongs to no single administration and to all of them (GAO, 2026a). The current example GAO chose is the Army’s augmented-reality headset program, which has burned through three acquisition efforts in eight years and is sending nearly 10,000 units into storage rather than into the field (GAO, 2026b).

None of that means administration policy does not matter. It means the unit of account is wrong. What accumulates over administrations is a portfolio, and what a portfolio delivers is a rate of programs that finished late and over budget, most of which were entered before the current occupant arrived.

What the administration says, in its own orders

The clearest statement of the problem is not from a critic. It is from the executive branch.

In January 2026, Executive Order 14372 barred major defense contractors from buybacks and dividends “until such time as they are able to produce a superior product, on time and on budget” (EO 14372). Read that condition literally and it states a fact about the present: the administration’s own position is that its contractors are not, at this moment, producing on time and on budget, and the order is the remedy.

In April 2026, Executive Order 14402 went further, making fixed-price contracts the default and pointing at a number of its own: a review of fiscal year 2024 spending “identified approximately $120 billion obligated on cost-reimbursement consulting contracts alone” (EO 14402), a contract form the order describes as providing little incentive to control costs. The order requires agencies to review and renegotiate their ten largest non-fixed-price contracts and to report semiannually to the Office of Management and Budget on every non-fixed-price contract they approve.

Notice what those reports will count: contract type. Fixed-price or not, justified or not, above or below the delegation threshold. Nothing in the order requires an agency to report what share of its contracts finished on time or under budget, because that figure has never existed to report. The administration is rewriting what the government buys and how it pays for it, and it is doing so without a published baseline against which to measure the change.

The dashboard that came closest is being switched off

The nearest thing to a public on-time and under-budget tracker was the Federal IT Dashboard, which displayed cost variance and schedule variance for major federal IT investments, with documented formulas for both.

It is closing. In a letter posted on the site, the federal chief information officer, Gregory Barbaccia, writes that the government is “taking steps to sunset this site, eliminating a costly, inefficient process and allowing agencies to focus on higher value activities,” with agencies to “pivot to a streamlined state that refocuses on statutorily required data” (IT Dashboard, 2026). GAO’s September 2026 assessment confirms the change from the other side: “In April 2026, OMB announced that it was refocusing the Dashboard to report on statutorily required data” (GAO, 2026c).

The last full year of data the dashboard carried told a similar story to the audits. Of the 24 programs GAO reviewed in 2025, the most over-budget had grown by $815.5 million, and the most delayed ran four years beyond its initial deployment schedule (GAO, 2025b). The dashboard was imperfect. It measured investments rather than contracts. It measured variance rather than outcomes. But it was the one public place where a reader could have watched, in dollars and months, whether the government’s largest technology projects were arriving on time, and it is being reduced to what statute compels.

The one clean denominator

Every so often, someone actually publishes a denominator, and it is worse than the argument.

A 2023 study in the proceedings of the Associated Schools of Construction conference used the Army Corps of Engineers’ own project database to examine 173 military vertical construction projects completed by the Fort Worth District between 2006 and 2020 (Tarver and Kramer, 2023). The result: an average of an additional 282 days “being added to the construction duration,” with 160 of the 173 projects finished behind schedule, three on time as originally scheduled, and ten ahead (Tarver and Kramer, 2023). That is 92 percent late, in one district, across a project type the corps builds constantly, over fourteen years that span three administrations.

It is not the national number. It is one district and one kind of project, and it says nothing about the $793 billion in annual contracts elsewhere. But it is what happens when somebody runs the query: the answer is not “most finish on time.” It is a long tail of lateness with a short head of exceptions, and the study’s own conclusion is that the agency had no uniform method for estimating how long a project would take in the first place.

What would make the question answerable

Three things, none of them exotic.

Publish CPARS in aggregate, stripped of contractor identity if necessary. The evaluations already exist and already contain cost and schedule ratings on a standard scale. What is missing is the tally, and the tally is not a technical problem. It is a policy choice, made in the name of source-selection confidentiality, and it is the reason nobody can answer the question at all.

Fix the reporting lag, or stop pretending the system is a real-time record. When one department registers its contracts an average of 421 days late, the aggregate is a picture of last year drawn by whoever got around to it.

Keep a baseline. If the government will not publish a completion rate, it could at least publish what share of major programs rebaselined in a given year, which is the honest leading indicator that “on time and under budget” is being redefined rather than met.

What the absence means

“On time and on budget” is doing rhetorical work in 2026 that no statistic is doing, and that is the thing worth noticing. It appears in an executive order as an unmet precondition for the right to pay dividends. It appears in the reform literature as the goal of every reorganization for thirty-five years. It does not appear anywhere as a measured quantity with a published trend, and the two artifacts that came closest, the IT Dashboard and the contract-level CPARS record, are respectively being retired and withheld.

This site carries a rule that applies here, and it is not a metaphor. A claim attributed to a source that does not contain it is worse than a claim with no source, because it looks checkable and is not. The same discipline applies to a percentage. Quote a figure and give it no denominator, and readers will assume the government measured something. Nobody did. The nearest published series counts unchanged programs in one department’s IT portfolio, one year at a time, in cohorts that overlap by less than half.

So the honest answer to the question is not a number and not a comparison. It is a statement about the record: the contracts exist, the evaluations exist, the late ones exist, and the share that finished on time and under budget has never been computed in public. What can be said with evidence is that the pattern is old, that it spans administrations, that the agency’s own orders treat failure as the status quo they inherited, and that the instrument built to watch it is being switched off.

A country that cannot state how many of its contracts arrived on time is not keeping the books on its own promises. Ask the question again in two years. If the answer has not changed, that will be the answer.

PRH | huffmanwrites.org | © Philip Huffman

Sources

• GAO, 2026a. “Weapon Systems Annual Assessment: Requiring Mature Technologies Could Enable Shift to Rapid Delivery,” GAO-26-108457, July 2, 2026. Source of the 12-year average time frame to deliver a capability and the finding that 18 of 40 programs entered the MTA pathway with immature technologies between 2018 and 2025.
• GAO, 2026b. “Weapon Systems Acquisition: Beyond Business as Usual — Using Leading Practices to Curb Waste and Save Billions,” GAO-26-109135, June 9, 2026. Source of “have consistently exceeded cost estimates and delivery schedules” since 1990 and of the Integrated Visual Augmentation System’s three acquisition efforts and nearly 10,000 stored units.
• GAO, 2026c. “IT Systems Annual Assessment: DOD Should Improve IT Fraud Risk Management Practices,” GAO-26-108596, September 28, 2026. Source of the 18 programs and $10.3 billion, the cost and schedule change counts and ranges, the three programs expecting to rebaseline, and the April 2026 Dashboard refocusing.
• GAO, 2026d. “A Snapshot of Government-Wide Contracting for FY 2025 (interactive dashboard),” GAO WatchBlog, May 5, 2026. Source of the approximately $793 billion in contract obligations.
• GAO, 2025a. “IT Systems Annual Assessment: DOD Needs to Improve Performance Reporting and Cybersecurity Planning,” GAO-25-107649, June 12, 2025. Source of the 24 programs and $10.9 billion, the 48-month maximum delay and 15-month median, and the $815.5 million largest cost increase.
• GAO, 2025b. “DOD Efforts to Buy and Maintain IT Systems Are Billions Over Budget and Delayed,” GAO WatchBlog, June 17, 2025. Source of “10 systems were on time and budget.”
• FAR 42.1501. “Subpart 42.15, Contractor Performance Information,” Federal Acquisition Regulation, current edition. Source of “CPARS is the official source for past performance information” and of the performance factors including cost forecasting and schedule adherence.
• DOT OIG, 2023. “DOT Faces Challenges in Meeting Federal CPARS Reporting Guidance,” U.S. Department of Transportation Office of Inspector General, ZA2023022, March 15, 2023. Source of the 421-day average registration time, the less than 26 percent compliance rate, and the summary of GAO findings on missing contractor performance information.
• IT Dashboard, 2026. “Open Letter to the American Taxpayer,” U.S. General Services Administration, retrieved October 9, 2026. Source of the chief information officer’s letter on sunsetting the site and refocusing on statutorily required data.
• Tarver and Kramer, 2023. Andy Tarver and Scott Kramer, “USACE Southwestern Division: Construction Durations for Military Vertical Projects,” EPiC Series in Built Environment, vol. 4, pp. 336–344, 2023. Source of the 173 projects, the 282-day average time growth, and the counts of 160 late, three on time, and ten early.
• EO 14372. “Prioritizing the Warfighter in Defense Contracting,” Executive Order 14372, January 7, 2026. Source of “until such time as they are able to produce a superior product, on time and on budget.”
• EO 14402. “Promoting Efficiency, Accountability, and Performance in Federal Contracting,” Executive Order 14402, April 30, 2026. Source of “approximately $120 billion obligated on cost-reimbursement consulting contracts alone,” the fixed-price default, and the semiannual reporting requirement.