What we measured, and how
Everything here comes from training.gov.au, Australia's national training register, which is public. There is no private data in this study and no provider was contacted for it.
- Population. We read 3,985 registered training organisations; 3,839 of them had a current registration on the day of the read, and those 3,839 are the population every figure on this page is computed on. The other 146 — pending, suspended, non-current or cancelled — are excluded: a provider that is not operating has an exposure that does not mean the same thing. The read covers every provider on the register's list of 17 August 2026; providers registered after that date are not in it. There is no sampling and no weighting anywhere: every figure is a direct read of the register, not answers from providers.
- Read in a single day, and that turned out to matter. Every figure here comes from one read of the whole register on 2 October 2026. An earlier version of this census was taken over eleven days, and comparing the two showed why that is not good enough: of the 662 providers measured on the first day with a superseded qualification, course or skill set on scope, 103 no longer had one eleven days later — we checked 25 of them one by one and in every case a scope entry for that product had simply reached its end date. The same effect halved the not-equivalent count. Products leaving a scope is exactly what this service watches for, so a census spread across eleven days is not one photograph, it is two.
- Only what a provider delivers. A scope also lists training packages — the container, not something you deliver — and some of those carry a superseded status. Counting them inflated an earlier version of these figures. Here, only qualifications, accredited courses, skill sets and units of competency are counted.
- Live scope only. We looked at what a provider lists on its scope today. Historical scope — products it used to deliver and no longer lists — is excluded, which is why our numbers are lower than a naive read of the register would give. Why that distinction matters.
- Status. "Superseded" means the register's own status field says superseded. We do not infer it and we do not date it ourselves.
- Equivalence. "Not equivalent" is the register's own flag on the replacement, published alongside it. It is not our reading of the two products.
- Equivalence, in one flag. What that flag decides — and why two identical-looking supersessions can mean very different amounts of work — is set out in our note on equivalent vs not equivalent. Checking a single product by hand is four steps.
- Date. The register was read on 2 October 2026. It changes daily, so this is one day's photograph, not a trend.
- Aggregate. No provider is named, ranked or identifiable here. The one extreme case we describe is already published anonymised.
Two definitions of "training product"
The register uses one phrase for two very different things, and the answer changes depending on which one you count. We publish both, because a study that quotes one number without naming its definition is not a study.
Strict: qualifications, accredited courses and skill sets — the things a provider advertises, enrols into and issues. Broad: every training product on scope, individual units of competency included. Units are superseded often, and usually in batches when a training package is updated, which is why the broad number is so much larger.
The gap is the whole point. Under the strict definition 0.3% of providers (13 of 3,839) have fifteen or more superseded items on scope; under the broad one, 33.3% do. Both figures are true of the same 3,839 providers on the same day. When someone quotes a number about this sector, the first question worth asking is which of the two they counted.
The numbers
Everything we measured, in one table. Each row can be re-derived from the register.
| Measure | Value |
|---|---|
| RTOs read | 3,985 |
| Of those, with a current registration — the population for every figure here | 3,839 |
| With at least one superseded qualification, course or skill set on live scope | 1,186 · 30.9% |
| With at least one whose replacement the register flags not equivalent | 149 · 3.9% |
| With fifteen or morestrict definition | 13 · 0.3% |
| Superseded qualifications, courses and skill sets counted in total | 2,414 |
| Median per affected provider | 1 |
| Most in a single provider | 81 |
| Median time since supersessionall 2,414 of them, each dated against the register | 7 months |
| Superseded more than five years | 52 products |
| Longest single product | 170 months |
| With at least one superseded product of any kindbroad definition — units included | 75.2% |
| With fifteen or more of any kindbroad definition — units included | 33.3% |
The shape of the distribution matters as much as the headline. Most affected providers have a small number — the median is 1 — and a few carry a lot: the largest single figure we found was 81 superseded qualifications, courses and skill sets on one live scope. 52 of the 2,414 products counted here have been superseded for more than five years.
How long the typical one has been sitting there used to be a separate measurement, because dating a product means reading its supersession record one by one and we could only do a subset. Now it is not: all 2,414 were dated in this read. The median is 7 months, and 7.8% — about one in 13 — have run past two years. Most of this is recent rather than decades of neglect, but the tail is real: 52 have been superseded for more than five years and the longest for 170 months.
The oldest one in the count
One product in this census has been superseded for 170 months — more than ten years — and is still listed on a live scope. Not hidden: published, in a public register, every day of it. That provider is not named here or anywhere else on this site.
If you want to see what one of these looks like up close, we have written up a different provider's case — a shorter one, at 128 months — with the register's own figures and no name.
The clock on top of the census
The count above says a superseded product is on a live scope. It does not say what the law then attaches to it. Measured on 2 October 2026 across the 3,628 ASQA-regulated RTOs with a current registration (counted on 20 August 2026) — and only those, because item 14 binds ASQA's providers and not those regulated by VRQA or TAC — 1,084 carry at least one superseded qualification, accredited course or skill set on live scope: 2,126 scope instances in all.
What happened on the day. Counted on 3 September 2026, two days out. Read again on the day itself, 5 September 2026: not one of them was still on an ASQA-regulated scope (26 scope entries remained, all of them under VRQA or TAC, which item 14 does not bind). Read again on 14 September 2026: 2 of them were back on an ASQA-regulated scope (CHC50121). The register does not close the scope entry when the computed cut-off arrives: it can leave it open for years.
Method. The cut-off is the day the replacement was included on the National Register plus one year, under s 14(1)(a)(i) of F2025L00355. We compute it; the register publishes every ingredient and never does the sum. How we read “included”. The register offers three dates a day apart for that event: the pair's mappingInformation.date (one day earlier on the three pairs we read), and the replacement's first releaseDate, currencyPeriod.startDate and usageRecommendation.startDate (the same day as releaseDate on the pair where we read them). We use the replacement's earliest releaseDate — its first release is its entry on the register — because, on the pairs we checked on 5 September 2026, it gives the same first day of the prohibition as ASQA's published transition example and as its extension notice for AMP20316. The NSW and Queensland subsidy lists end their funding one day earlier; they are not the instrument. If the mapping date were the inclusion, the prohibition would start one day before the date we show — an error that would run against the provider, not for it, so we treat the field as checked on three pairs and not on the whole index: that count is still pending, and if it moves the field, every page moves with it. Two things this measurement deliberately does not claim: it is not a count of providers doing anything wrong — ASQA can approve otherwise, for one provider or for a whole class — and it says nothing about learners already enrolled, who have no fixed deadline at all. The full working, product by product →
What these numbers do not say
Three things this study is not
Not a judgement about any provider. A superseded product on scope is a fact of the register. Transition arrangements differ by product and by circumstance, and assessing them belongs to the regulator, not to us.
Not enrolments. Scope is what a provider is registered to deliver, not who is sitting in it. A superseded product on scope may have no learners in it at all. How many learners are actually in one is visible only inside a provider's own student management system — which is exactly why our enrolment cross runs in your browser and never sends that file to us.
Not our opinion on equivalence. Every "not equivalent" here is the register's published flag on the replacement. We copy it; we do not assign it.
A correction to this page
This section used to say the opposite
Until 7 August 2026, this study said that what a qualification is made of was not published at all, and that the unit-by-unit comparison between a superseded product and its replacement could therefore not be derived from the register — by us or by anyone. That was wrong. It was wrong because of how we asked, not because of what the register holds: training.gov.au does publish the list of units that make up a qualification, and we had been looking for it in the wrong place.
We have left the mistake written down here instead of editing it out. A study that will not correct itself in public is worth less than the day it was published, and this one was wrong in the direction that happened to suit us.
What the register publishes about composition
A qualification lists the units inside it: each unit's code and title, whether it is core or elective, and whether that unit is itself still current. Superseded qualifications keep their list too, which is the part that matters — it means an old qualification and its replacement can be set side by side without anyone reading a document.
We checked this against qualifications taken from the live scopes of providers spread across this census, not from a convenient list: all 50 current qualifications we tried returned their unit list, and so did all 7 superseded ones — averaging 48 units for the current ones and 55 for the superseded. Anyone re-running this should know the register returns errors if it is read too fast; without retrying, coverage looks far worse than it is, which is how a check like this ends up producing a confident wrong answer.
Why subtracting the two lists overstates the change
Comparing the two lists directly is easy and misleading, because most of what appears to vanish is the same unit under a new code. Taking HLT54115 against HLT54121 and working it through in full, the raw subtraction showed 29 units leaving and 26 arriving. Once each unit's own supersession chain was resolved, the picture was: 13 unchanged, 14 renumbered with the change flagged equivalent — paperwork — 13 renumbered and flagged not equivalent, 2 genuinely new and 2 genuinely dropped. The work sits in the third and fourth of those, not in the fifty-five the subtraction implied.
Where a unit has been renumbered more than once, the chain only counts as equivalent if every step in it is flagged equivalent. One step that is not is enough to make the whole route not equivalent, so a comparison that stops at the first hop reads as reassuring when it should not.
What we still could not see
One thing, stated narrowly so it is not overstated a second time. The register publishes which units make up a qualification, but not the packaging rules — how many electives must be chosen, and from which groups. Those rules live in the Companion Volume documents, which the register does publish as downloadable PDFs. A PDF is a document, not a field: it cannot be queried, compared between versions or watched daily without building something to read it first. We have not read those documents and we make no claim about what is in them.
The other limit is unchanged, and it is not the register's to fix: it cannot see how far your own learners have got. That exists in your student management system and nowhere else.
Limits
This study describes what the national register publishes. It is not a judgement about any centre: transitions have their own arrangements and circumstances, and weighing them is the regulator's role.
It counts providers we could read with a current registration on one day; the register changes daily, so a re-run gives slightly different numbers. It counts live scope only, so it undercounts anything a provider has already removed. It says nothing about learners, delivery or quality, none of which the register publishes. And it names nobody: every figure here is an aggregate.
Read from training.gov.au, Australia's national training register, on 2 October 2026. 3,839 RTOs with a current registration; live scope only, historical scope excluded. Qualifications, accredited courses and skill sets: 1,186 providers with at least one superseded (30.9%), 149 with at least one replacement flagged not equivalent (3.9%), 13 with fifteen or more (0.3%); 2,414 superseded products counted in total, median 1 per affected provider, most in a single provider 81; median 7 months since supersession, 52 superseded more than five years, longest 170 months. Counting every training product on scope with units of competency included: 75.2% with at least one, 33.3% with fifteen or more. No provider is named, and no provider was contacted for this study. Unit composition checked separately on 7 August 2026 against qualifications drawn from live scopes across the same census: 50 of 50 current and 7 of 7 superseded qualifications returned a unit list, averaging 48 and 55 units; the HLT54115 to HLT54121 comparison was resolved unit by unit through each unit's own supersession chain.
Check your own scope
The same read this study is built on, run on one RTO code: your live scope, what the register says about each product, where it points next, and the date new enrolments into it stop. It is free and every line is checkable against the register.
Straight link, if you already know your code: registerradar.com/?rto=YOUR-RTO-CODE — or just run the check here.
Journalists and sector analysts: the underlying counts, the definitions and the day's read are yours on request — write to contact@registerradar.com and we will send them as they came out.