The year that is running, not the last one that closed

FY2027 at 361 days out

Four figures a review would start from on this date, each labelled with how far its source actually reaches. The first two are the whole Department's DIRECT execution, from the account files — reimbursable work, paid back by a customer, is shown beside them and not added in. The second two are contract actions only, which are 33.9% of obligations — the account files carry no date, so anything about timing can only be read from them.

The September figure is a projection: each sub-agency’s observed whole months scaled by how September has related to that same window in its own prior years, and the range is the observed minimum and maximum of that ratio, not a confidence interval. It is what the year end looks like if each organisation behaves as it has, and it will be wrong for any that does not.

The Department-wide position, FY2026

The whole of execution, not a subset — every appropriation, every account. Read top to bottom: the first three rows assemble the resources, the total is the footing, the next two are the flows against it. Every line is period-to-date at FY2026P10.

This is a position, not a curve, and it cannot be made into one. File A and File B each publish one submission per fiscal year in this warehouse — FY2026 at FY2026P10, every closed year at P12 — so there is no month-by-month Department-wide series in these sources to draw. The timing section below is built on contract action dates because an action carries a date and an account submission does not; it covers 33.9% of obligations, and that is stated there rather than left to be inferred.

Controls SBR-01 and SBR-02 assert that obligations plus unobligated balance equal total budgetary resources, and that the resource components sum to that same total. Both foot to within 0.001% for every year shown.

How long the money lasts

Direct obligations by period of availability, read off the beginning and ending periods on each account. This is the split the date at the top of this page is about. Reimbursable work is excluded.

Execution by program year

File B, split by the year the money was appropriated for. A fiscal year's execution is not that year's money: FY2026's obligations include prior-year procurement and research money still being obligated, expired annual money taking adjustments, and no-year money, beside FY2026's own appropriations. Keep to one program year, then drill from component to object class; each row's rate is its own obligations over its own resources.

A position at each submission, not a monthly curve. The warehouse holds one File B submission per fiscal year (FY2026 at FY2026P10), so what moves across these charts is a program year’s life — the same money in its first, second and third fiscal year — not the months of one year. The program year is the beginning of the period of availability; control POA-01 re-derives it from every Treasury account symbol and POA-02 asserts the split foots to the Statement of Budgetary Resources.

When contract money moves

Cumulative CONTRACT obligations by day of the fiscal year — 33.9% of Department obligations in FY2025, not the whole of execution. It is here because an action carries a date and an account submission does not. The FY2027 line stops at the reporting frontier — 2026-04-30, 7 whole months in — and the gap between the two markers is the reporting lag, not a fall in spending.

This curve is the Department as a whole against the calendar year. Fund distribution and the execution timeline takes the same contract file and cuts it by fund holder — WHS, MDA, SOCOM and the rest, which appear in no other source here — and reads every year against the continuing resolutions, lapses and appropriation acts the Department actually operated under.

The FY2026 file carries actions dated as late as 2026-08-04, but it is substantially complete only to 2026-04-30: October through April carry between 280,000 and 400,000 actions a month, and everything after the frontier comes to 75,264 actions worth $9.0B. Every figure on this page measures the live year to the frontier and every prior year to the same point. Taking the last dated action as the extent of the file instead put three near-empty months into FY2026’s total and none into anyone else’s — which showed every organisation running behind its own norm when all of them are running ahead. Control TIME-04 asserts the frontier and publishes what it excludes.

Signals worth asking about

Each one compares a category with its own prior years, never with a Department-wide average — these categories differ by three orders of magnitude in size and a shared threshold would only ever select the largest of them. A signal is a question, not a finding.

None of these observes impropriety, and several of the largest are certainly ordinary: a multiyear definitisation, an exercised option or a supplemental lands as one very large action with nothing unusual in it beyond its size. The deviation is a robust z — the distance from the median of the category’s own years, scaled by their median absolute deviation, floored so that a category with a nearly flat history cannot produce an arbitrarily large number, and capped at 99. A capped value means far outside its own history rather than a measurement. Control TIME-02 refuses a signal computed against fewer than three of a category’s own years.

Who is executing, and how much of it is left to September

Each sub-agency against its own norm for the same whole months, with the September and final-five-day dependence its own prior years show. Ordered by size, because a pace percentage on a small organisation is not comparable with one on the Navy.

What a year end actually buys

The individual actions, with the description the contracting officer wrote on them.

Scope, before any of it counts

File A carries five agency identifier codes. Four are the Department; 011 is the Executive Office of the President. Every figure on this page uses the Department scope only.

Control SCOPE-01 asserts that no Department-scope figure includes agency code 011, and runs on every load.

Execution detail

File B at the grain it is reported at: 24,352 rows for FY2026 across 856 Treasury accounts and 35 object classes. Expand a row to open the level beneath it.

The FY2026 submission carries no program activity name on any row — the column is null throughout and the activity is identified by its reporting key instead. The key is shown; no name is borrowed from another year’s row that happens to share it. Where an account holds several keys the file repeats the account’s object-class figure against each one rather than splitting it, so 1,505 repeated rows are counted once; summed as published they run about a third above File A. The break is set out on linkage.

Where obligated dollars sit

An obligation is a binding reservation, not a payment. File B splits it by USSGL account into undelivered orders, delivered orders, and what has already been outlaid. Direct obligations only.

A rising undelivered-orders balance means more of the year’s obligations are still ahead of delivery. It is a pipeline measure, not a performance measure: multi-year procurement and construction accounts carry large undelivered balances by design.

A cross-system reconciliation that does not tie

File A and File B are separate submissions of the same execution, at different grain and read here at the same submission period. They should agree closely. In 1 of 6 years they do not, and that variance is published rather than hidden. Both columns are direct and reimbursable together, because that is what each file publishes and what reconciles; the direct split is above.

The File B column is not a plain sum for FY2026 — see the note on the explorer above.

Execution rates across the window

Obligation and outlay rates against total budgetary resources. The in-progress year is marked and must not be read beside closed years as if it were one.