INSIGHT · RISK & FINANCIAL VISIBILITY

Construction insolvencies:
July held at 292.
August went from 59 to 826.

In August 2026, the second month of FY2026-27, 826 Australian construction companies entered external administration or had a controller appointed for the first time (ASIC Series 1); Series 2, which counts every appointment, reads 958. That is our recount of ASIC's workbook released 28 September 2026, with data to 13 September, and it matches ASIC's own tables. ASIC's note says August includes 542 related companies of one large property development and construction group. ASIC's tables still shade August as provisional. July 2026 read 292 in both the August and September workbooks.

Analysed 28 September 2026 · ASIC Series 1 and Series 2 workbooks released 24 August 2026 (data to 9 August 2026) and 28 September 2026 (data to 13 September 2026), recounted from record-level entries · ASIC INFO 80, updated February 2026 · all pages read 28 September 2026.

Key takeaways

• August 2026: 826 companies, 958 appointments: Construction division, ASIC Series 1 and Series 2, workbook released 28 September 2026 with data to 13 September 2026. Our recount from the workbook's 75,056 record-level entries, which matches the Construction column of ASIC's monthly tables 1.4.3 and 2.4.3.
• One group sits inside the August figure: ASIC's data note says August includes appointments involving 542 related companies of a large property development and construction group. Because the statistics count individual companies rather than corporate groups, ASIC states these appointments "have materially affected the results for the month".
• July did not move between releases: Construction July 2026 read 292 companies (Series 1) and 383 appointments (Series 2) in the workbook released 24 August 2026, and again in the one released 28 September 2026. For this cell, the five weeks between the two releases added nothing.
• The August jump is new days, not late lodgements: The 24 August workbook showed August at 59 companies because its data stopped at 9 August. The 28 September workbook still reads 1 to 9 August at exactly 59. The difference is the rest of the month arriving, not earlier days being revised.
• We would store the month with its series, cell, release and note: A monthly insolvency count is reusable only if it travels with its series, the industry label it was counted under, the days it covers, the workbook's release and data-to dates, and any publisher data note. That is our system-design recommendation, not a reporting standard.

The months, as the
28 September workbook reads them

Construction division, recounted from record-level entries in the workbook released 28 September 2026 (data to 13 September 2026). The same months of FY2025-26 are read from the same file, so both years sit at the same cutoff.

Our calculation. Series 1 counts a company the first time it enters external administration or has a controller appointed; Series 2 counts every appointment, including subsequent and transitional ones. The month is the appointment's effective date, which ASIC says is how the data is collated, rather than the lodgement date. The status column reports ASIC's blue shading of provisional figures in tables 1.4.3 and 2.4.3.

• A like-for-like September, for context: The first 13 days of September 2025 read 111 companies and 146 appointments in the same file, against 133 and 247 for the first 13 days of September 2026. Neither is a monthly total, and the 2026 figures are provisional.
• Series 2 ran well ahead of Series 1 in early September: Of the 247 construction appointments dated 1 to 13 September 2026, 114 are not the company's first; 69 of those 114 are recorded as "Controller appointed (except receiver or managing controller)". That is a count of appointments, not of newly affected companies, which is why ASIC's INFO 80 prefers Series 1 when explaining trends.

01

July 2026:
the month that did not move.

ASIC releases Series 1 and Series 2 weekly, two weeks in arrears, and says appointments with start dates less than one month old are considered provisional; it updates the dataset for new appointments because registered liquidators may lodge appointment documents up to one month after the start date. The 24 August workbook's tables still shaded July 2026 as provisional. We compared the two workbooks cell for cell.

Our calculation: each workbook recounted from its Data set sheet (Construction division; Series 1 and Series 2 flags; month from the appointment's year-month field), then checked against the month totals in ASIC's tables 1.4.3 and 2.4.3, which agree in both files.

• What ASIC's documentation says can move: INFO 80 gives statutory lodgement periods, including within one business day after appointment for voluntary administrations, and within 14 days for court liquidations, creditors' voluntary liquidations and controller appointments. It adds that the lodgement may be late and that paper forms may not yet be processed. Changes are reflected while statistics are provisional; after that, ASIC states it "does not amend the published statistics".
• Something did move in July, outside construction: Across the whole economy, the July 2026 record set changed slightly between the two files: one appointment row was removed, and one appointment's industry moved from "Unknown" to Manufacturing. INFO 80 explains that "Unknown" means the Form 505, which carries the industry code, had not been lodged. The industry cell is provisional too, not only the count.
• One cell, one pair of releases: This is a single observation. It does not show that a closed month never moves. It shows that whether a month moved is something to check between two named releases, and to record as checked, rather than to assume in either direction.

02

August's jump is the rest of the month,
and one group inside it.

The 24 August workbook was not wrong about August. It covered nine days of it. Re-read in the 28 September workbook, those nine days still hold 59 companies and 82 appointments. The other 767 companies in the August Series 1 figure have effective dates from 10 to 31 August, days the earlier file had not yet reached.

• ASIC's own note on the month: The workbook's Contents sheet states that the August 2026 statistics "include external administration and controller appointments involving 542 related companies" of "a large property development and construction group", and that "This should be taken into account when comparing August 2026 with earlier months, particularly the results for New South Wales and the Construction industry."
• What the records show, and what they do not: 549 of the 826 construction Series 1 entries for August carry one effective date, 25 August 2026, and 544 of those 549 give New South Wales as the principal place of business. The Data set has no field that identifies a corporate group, and ASIC publishes no August figure without the group, so we do not construct one. Any such figure would be an estimate, not ASIC's statistic.
• Why the unit matters: INFO 80 describes Series 1 as the number of companies entering external administration or controller appointments for the first time. A group of related companies enters as many companies, not as one. A monthly company count therefore also reflects how the entities involved were structured, which ASIC's note says in its own words.

03

What a monthly figure
has to carry to be reused.

Someone who wrote down "August 2026: 59" on 24 August and "August 2026: 826" on 28 September recorded two true readings of the same cell. Stored without context they look like a data error or a sudden surge. Stored with it, they are two cutoffs. Our earlier FY2025-26 analysis makes the year-level version of this point; the monthly version adds four things a year hides.

• A partial month looks like a whole one: Unless the data-to date is stored, 59 reads as August's count rather than as nine days of it.
• A publisher note changes what a comparison means: ASIC asked readers to take the group into account when comparing August 2026 with earlier months. A figure copied out of a table without its note loses that instruction.
• The industry cell is a lodged value: INFO 80 says industry comes from the ANZSIC 2006 code on the Form 505 lodged by the external administrator or controller. The workbook's division label is "Construction"; a subdivision, a different label or a record still coded "Unknown" gives a different number.
• The month is the appointment month: ASIC collates by date of appointment, not by lodgement date. That is why a month that has ended can still gain entries while its latest appointments are under one month old.

The FY2025-26 analysis

Why this lands in your records,
not in the headline.

Monthly insolvency figures get copied into documents that outlive the week they were read in: a board pack, a credit committee paper, a risk section in a funding submission, a management commentary. Two documents a month apart can cite ASIC correctly and still disagree, because they read different workbooks. The question that follows is always the same, and it is operational rather than statistical: which release did each one use?

• The reconciliation is cheap only if the release was stored: With the workbook's Released and Data to dates on the record, the difference between 59 and 826 explains itself in one line. Without them, someone has to rebuild which file was open on which day, and the statistics page links only to the current workbook: the 24 August file's address no longer resolves.
• A qualified comparison should stay qualified: ASIC's August note is a caveat on comparisons, not on the count. A monthly series stored as bare numbers drops it silently, and every later chart built on that series inherits the drop.
• "No change" is also a reading: July's 292 was the same in both files. Recording that it was re-checked against the 28 September workbook is what lets a later reader trust the figure without re-opening two spreadsheets of about 16 MB each.

What we would put in a system

For each external benchmark figure a business stores, we would test for five fields.

Where agents help here is narrow. A finance agent can watch for the weekly release, archive the workbook, recount the cells a business tracks, and draft a note of what changed since the last stored reading, with the data-to date and any publisher note attached. A person confirms the stored figure before it goes into a report. No agent decides what a figure means for any counterparty, and the workflow makes no assessment of any company.

• One source, two numbers: The FY2025-26 construction count and why its series and cutoff decide it.
• Finance & Cashflow Agent: Drafts the refresh note when a new release lands; a named person confirms before anything is relied on.
• Reporting: Board and management reports that cite each external figure with its release.
• Group Finance & PM Platform: Holds one versioned benchmark record that every entity in a group reports from.

Questions worth asking
of your own records

For a construction group's finance lead: when your board pack quotes an ASIC construction insolvency figure for a month, does it say which workbook and data-to date it came from?

Quoted from the 24 August 2026 file, August read 59; from the 28 September 2026 file, 826. Both are ASIC's numbers. Without the release date, the next pack cannot say which one it is updating.

For a supplier or distributor: does your credit committee paper say whether it tracks Series 1 or Series 2?

For construction on 1 to 13 September 2026 the two read 133 and 247. They count different things, and which one a policy tracks is a choice to write down once. This is a record-keeping point, not credit advice.

For a developer or builder writing a risk section: if you cite August 2026, does the citation carry ASIC's note about the group?

ASIC asks readers to take 542 related companies into account when comparing August with earlier months. A citation that drops that sentence passes on a comparison ASIC itself qualified.

For anyone who stored a July 2026 construction figure in August: does it need updating?

Not on these two releases: July read 292 and 383 in both. That was worth checking rather than assuming, and worth recording as checked against a named release.

For an analyst filtering by industry: which label did you filter on?

The workbook's division label is "Construction". A record coded "Unknown" sits outside it until the industry form is lodged, and one July 2026 appointment elsewhere in the economy changed industry between these two files.

What this analysis
does and does not show.

What to watch next

ASIC publishes a cadence rather than a calendar of dates, so this list is short.

• The weekly release: ASIC states that it releases Series 1 and 2 weekly, two weeks in arrears. The statistics page links to the current workbook only, and the file carries its own Released and Data to dates.
• August's provisional window: INFO 80 treats appointments less than one month old as provisional. ASIC's tables in the 28 September 2026 workbook still shade August 2026 as provisional, so ASIC's own rule allows the August count to change in a later release.
• September 2026: The 28 September workbook covers 1 to 13 September. The rest of the month arrives through the weekly releases, on ASIC's stated two-weeks-in-arrears basis.

Bring us the last report
that quoted an insolvency figure.

Send us the board pack, credit paper or risk section that cited an ASIC construction count. We will show you which release it most likely came from, what that release's data-to date was, and what the same cell reads today.

General information about published regulatory statistics. Not financial, credit, legal or investment advice, not a forecast, and no assessment of any company, group or counterparty. Figures are national (Australia), from ASIC workbooks released 24 August 2026 and 28 September 2026, read 28 September 2026; recent months are provisional.

Send us the report · Finance & Cashflow Agent

Sources

Suggested citation: ASIC Insolvency statistics, Series 1 and Series 2 workbook released 28 September 2026 (data to 13 September 2026), Construction division, recounted at record level by Cyberate, read 28 September 2026. Derived rather than quoted: the monthly counts, the 1 to 9 August re-read, the 767-company difference, the 549 entries dated 25 August (544 in New South Wales), the 114 September appointments that are not a company's first, and the comparison with the workbook released 24 August 2026.

• ASIC - Insolvency statistics: No visible last-updated line; page metadata shows modified 28 September 2026 · read 28 September 2026 · government regulator, statistics release page
• ASIC - Series 1 and Series 2 workbook, released 28 September 2026: Released 28 September 2026, data to 13 September 2026 · read 28 September 2026 · record-level dataset, recounted
• ASIC - Series 1 and Series 2 workbook, released 24 August 2026: Released 24 August 2026, data to 9 August 2026 · archived copy read 28 September 2026 · ASIC has replaced the file, so this links to the statistics page
• ASIC - INFO 80: How to interpret ASIC's corporate insolvency statistics: Updated February 2026 · read 28 September 2026 · regulator interpretation guidance