When government work is actually released — and the two months it nearly stops

Half the year arrives in four months

Measured inside each complete calendar year separately, the release curve has the same shape every time. It starts at its lowest in January, climbs through autumn, peaks somewhere between August and November, and falls away in December. The second half of the year carries between 61 and 72 per cent of everything released.

In the most recent complete twelve months of the corpus the extremes are concrete: 1,587 listings released in July 2026 against 739 in February. If you bid in a market that publishes at that rate, February offers you fewer than half the opportunities July does — from the same buyers, doing the same work.

The 2025 curve is the sharpest of the three and the 2023 curve the flattest. We are deliberately not reading a trend into that. Comparing the intensity of the seasonal swing between years requires the year totals to be comparable, and in this corpus they are not — older years hold fewer records because of how long the archive retains them. What survives that problem is the shape inside each year, and the shape is consistent.

Why we nearly did not publish this

The finding above has an obvious alternative explanation, and it is not seasonality. If old listings drop out of an archive over time, then inside any single year the January records have had eleven more months to disappear than the December ones. That alone would manufacture a rising curve from nothing. Any dataset built from an archive will show it, and most published “tender season” claims never check.

The check is to compare cohorts of different ages. Retention acts on age, so if the shape is retention it must be much weaker in records that are young. We took the January-to-July shape of 2026, where the listings were between one and seven months old when the snapshot was taken, and set it against the same months of 2025, where they were thirteen to nineteen months old. If age drove the curve, the young cohort would be visibly flatter.

It is not. The two shapes rank-correlate at 0.9286, and the largest disagreement on any single month is 2.23 percentage points. A cohort barely touched by retention has the same curve as one that has had a further year to erode. Whatever is producing the shape, it is not the archive forgetting things.

We set that threshold before running the test and wrote it into the script: rank correlation of at least 0.80 and no month differing by more than 4 points, plus reproduction across three separate years. Had it failed, this page would not exist rather than exist with a caveat.

What drives the calendar

The two features of the curve line up with two facts about how Australian government spends. The financial year begins on 1 July, and in the most recent complete twelve months July was the busiest month of all at 1,587 listings. Budgets are allocated, programmes start, and the work that was approved in May gets published in July.

The second feature is the summer trough. January sits at 3.1 to 4.5 per cent of the year across the three measured years — the lowest month every time — with February close behind. Australian government substantially closes between Christmas and the end of January, and a tender published then would run its response window through the period when nobody is available to answer it.

The consequence for a bidder is a calendar, not a strategy. December and January are when the reusable half of a submission gets built: the capability statement, the referee list, the accreditations, the past-project evidence. February to April is when you have spare capacity and the market does not. From May the volume climbs and does not stop until December.

Your state may not follow the national curve

The national curve is an average of eight quite different state curves, and one of them barely moves. South Australia published between 40 and 104 listings in every month of the measurement window — a swing of about two and a half times, driven mostly by a single quiet April rather than a season. If you bid only in South Australia, the tender calendar is close to irrelevant to you.

The large states swing hardest. New South Wales ran from 154 listings in January to 383 in October. Victoria ran from 89 in February to 272 in July. Western Australia ran from 138 in December to 291 in July. In each of those markets the busy month offers between two and three times the opportunities of the quiet one.

Queensland and Tasmania sit between the two patterns, and the Australian Capital Territory is too small a market for its monthly counts to say anything reliable — 274 listings across a year means most months carry fewer than thirty, and a single large programme moves the shape.

How to use a tender calendar

The practical value of knowing the shape is not that you can find work in February that is not there. It is that you can stop treating a quiet month as a failure of your search and start using it. The reusable half of every bid — who you are, what you have done, the certificates that prove it — takes weeks to assemble and never gets faster by starting it inside a live window.

The second use is capacity planning. If about half the year’s opportunities land between August and November, and the median response window is 24 days, then the constraint in spring is not discovery but your ability to write. Suppliers who arrive at August with current referees and a maintained capability statement can answer three tenders in the time it takes an unprepared competitor to answer one.

The third is expectations. A quiet January is not the market rejecting you and it is not your search being broken. It is 3 per cent of the year.

• December to February: build and refresh the reusable material. Volume is at its floor.
• March to May: the climb starts. Register on the portals you are missing.
• June and July: the financial year turns. July was the busiest month of the last twelve.
• August to November: about half the year’s work. Writing capacity is the constraint, not discovery.
• If you bid only in South Australia, ignore all of the above — that market is close to flat.

How we measured

The corpus is a Stipple research dataset of 33,064 publicly listed tenders observed on 29 August 2026, of which 26,073 positively carry an Australian or New Zealand region tag and 23,401 of those carry a release date. Month shares are computed inside each complete calendar year separately, never pooled across years, because the corpus holds far fewer old records than recent ones and pooling would measure retention rather than season.

The retention test is the reason this page exists. Test A compares the January-to-July month shares of 2026, whose listings were 1 to 7 months old at the snapshot, against the same months of 2025, whose listings were 13 to 19 months old, on the reasoning that retention acts on age: Spearman 0.9286 with a largest single-month gap of 2.23 percentage points. Test B requires the twelve-month shape to reproduce across 2023, 2024 and 2025 independently: pairwise Spearman of 0.9632, 0.9107 and 0.9231. Both thresholds were fixed in the script before the test ran, and both are recorded in the output artefact alongside the result.

Two exclusions. The measurement window runs 25 August 2025 to 24 August 2026, so August appears as two part-months at either end; the monthly count chart excludes both rather than showing them as artificially quiet. And no figure on this page compares one year’s volume to another’s, because year totals in this snapshot reflect archive retention and not market size. Coverage is one research snapshot, and every share is a share of these listings.