Monday 19 October 2026 • By Timothy, CPA — Managing Director, Professional Financelink (PFL)

A funding stream splitting into three channels, the middle one flowing freely while the outer two silt up and narrow

StewartBrown's FY26 Home Care Numbers: Net Admin Cost Up About 60% Per Client, and Roughly a Quarter of Package Funding Unused

StewartBrown's FY26 survey landed last week. The home care result isn't in the direct care line — it's in the two numbers either side of it, and only one is a pricing problem you must wait on government to fix.

StewartBrown released its FY26 Aged Care Financial Performance Survey at its Finance Forum on 14 October. Most coverage went to residential care, where the numbers are genuinely alarming. I want to spend this post on the home care numbers instead, because this is the first full-year survey to include Support at Home — eight of FY26's twelve months ran under the program, which began on 1 November 2025 — and because two of them are the kind of figure a finance lead can act on before Christmas.

The survey covers approximately 90,000 Support at Home packages — a large sample, but a voluntary one. It surveys participating providers rather than censusing the sector, so read it as a strong indication of direction, not a population statistic.

$16.85 Net administration cost per client per day, FY26 — up about 60% from $10.52. June quarter: $19.10.	76.0% Package utilisation, FY26 — down from 88.2% in FY25. June quarter: 72.2%.
−$0.78 June quarter operating result per client per day. The full year still finished at a $0.26 surplus, against $3.77.	5.14 hrs Care delivered per client per week, down from 5.35 — less service through the same overhead.

One framing correction before going further, because it has been reported loosely. The June quarter deficit is not the first negative quarter for home care — the March 2026 quarter was, at −$0.46 per client per day, reported back in August. June is the second consecutive negative quarter, and a deeper one. That matters: a single bad quarter during a programme transition is noise. Two in a row is harder to dismiss — even though June-quarter utilisation, at 72.2%, was up on the 62.2% reported for March. Utilisation is recovering from a very low base, not back anywhere near FY25. (The releases draw on separate samples, so treat quarter-on-quarter moves as indicative.)

Two numbers, two completely different problems

It is tempting to read $16.85 and 76.0% as one story called "Support at Home is underfunded". They are not the same problem, do not have the same owner, and only one needs a policy change to move.

The administration number is a structure problem. It is what it costs to run the organisation around the care, expressed per client per day. If it rose about 60% while hours of care fell, the overhead did not get more expensive so much as it got spread across less delivered service.

The utilisation number is a revenue problem with at least three different causes, and finance is usually the only function that can tell them apart. More on that below.

The two call for opposite responses. A structure problem is addressed by changing the shape of the overhead; a utilisation problem is addressed by changing what happens in the first eight weeks of a client's package. Organisations that treat both as "wait for better pricing" tend to do neither.

Why the admin line moved: a revenue line that stopped existing

StewartBrown has been consistent about the mechanism through the transition. Under the old Home Care Packages arrangements, providers charged a package management fee alongside care management. Under Support at Home, that fee is no longer available. The administrative work it funded did not disappear with it.

Add the second effect — lower service volumes meaning fixed overheads are recovered across fewer delivered hours, at prices set when the programme started — and you get a line that looks like a cost blowout but appears to be largely a revenue problem landing on a cost base that barely moved. The half-year figures point the same way: gross administration costs went from $21.51 per client per day in the September 2025 quarter to $21.85 for the six months to December, while package management revenue fell from $11.07 to $4.85 in the December quarter. StewartBrown's Stuart Hutcheon has also called care management the trickiest part of the new model for providers, because in principle there is less funding for it.

The distinction matters at board level. Present $16.85 as "our admin costs are out of control" and you will be sent away to cut administration. Present it as "the fee that funded this work was removed and the work remains" and the conversation becomes which of that work is still required, which can be absorbed into direct service delivery, and which only ever existed to administer a funding mechanism that is gone.

Utilisation is not one number — it is at least three

Twenty-four per cent of package value going unused is the figure I would be taking apart this week. In the organisations I see, a sector-average utilisation gap typically breaks down into at least three causes that look identical in the general ledger and are completely different problems:

1. Allocated but not wanted. The client's assessed package is larger than the service they actually accept. Nothing is broken; the revenue was never going to be there. This portion should be in your forecast as a permanent discount, not chased.

2. Wanted but not staffed. The client would accept more service and the roster cannot supply it, often in a specific geography or a specific support type. This is a workforce and scheduling problem showing up as a revenue problem, and it is the most expensive of the three because you are carrying the overhead of a client you are under-serving.

3. Delivered but not claimed. The service happened, the staff member was paid, and the claim is late, rejected or sitting unbilled. This is pure finance exposure — the cost is recognised and the revenue is not — and it is the one where a report, rather than a strategy, fixes it. It is the same mechanism I wrote about for NDIS unbilled services ageing, applied to a different funding stream.

A single sector utilisation percentage tells you nothing about which of these you have. Splitting your own number into those three buckets, by client cohort and support type, takes a few days and changes the next twelve months.

Four lines worth reporting monthly, not quarterly

The quarterly survey cycle is too slow to manage by. Four lines, every month, on one page:

One denominator, used everywhere. Pick occupied client days and express every cost and revenue line against it. Mixed denominators make a sector comparison impossible to use.

Net administration cost per client per day, split into rate and volume. If this line rises, you need to know immediately whether the overhead grew or the client days shrank beneath it. Those are different conversations and the single number hides which one you are having.

Utilisation by cohort, in the three buckets above. Reported with the newest cohort separated out, because onboarding lag will otherwise make a healthy book look like a deteriorating one.

Delivered-but-unclaimed service value, aged. Measured from date of service, not date of claim. If this has a tail beyond 60 days, it is your most recoverable number on the page.

Where AI genuinely helps here — and where it doesn't

This is a data-shape problem, which is where current tools earn their keep. Three uses hold up.

The first is normalising time records. Rostering, payroll and claiming systems rarely describe a visit the same way; AI is good at proposing the mapping between them and flagging the ones it isn't sure about.

The second is exception detection: "which clients' utilisation fell in each of the last two months" is a question nobody asks of 600 clients by hand, and it has an unambiguous answer.

The third is drafting — the first sentence of a variance explanation, not the explanation itself. I covered the boundary in detail in AI and variance commentary: the tool describes what moved, a person owns why it matters.

Where it does not help is deciding which utilisation bucket a given client belongs in. That requires knowing whether the care team tried to increase service and was refused, and that lives in a conversation, not a dataset. We build this kind of unit-cost and exception reporting for clients, and the useful part is never the model — it is agreeing the denominator and the data mapping first, which is bespoke to how each provider records a visit.

A note on AI tools and your data: client schedules, care notes and payroll exports are exactly the files that get pasted into an AI assistant to "just tidy this up". Before any of it goes into a tool, confirm whether the vendor retains your inputs or uses them to train its models. The safest posture is a tool where your data isn't retained for training — and de-identified client references even then.

This post is general commentary based on publicly available information and does not constitute legal or tax advice. Survey figures are as reported by StewartBrown and sector press — confirm against the source report before relying on them. Always seek independent professional advice before acting.

Can you split your utilisation gap into its three causes this month?

Most home care providers can report the percentage and not the reason — which makes it impossible to act on. PFL provides senior-level outsourced finance, management reporting, and AI automation for Australian NFP, NDIS, and SME organisations, including unit-cost reporting built on a denominator your board can actually compare.

Talk to PFL →

Timothy, CPA is Managing Director of Professional Financelink (PFL), providing senior-level outsourced finance, management reporting, and AI automation for Australian NFP, NDIS, and SME organisations. 20+ years in finance leadership across NFP, NDIS and SME.

Sources

Tomorrow: the September quarter BAS is due on 28 October. A practical way to put AI across your GST coding as a second pair of eyes — and the one step you cannot delegate to it.

Search This Blog

Finance Intelligence