Western Australia built a second road to an unrestricted electrical licence. The unit-level data shows how little of it runs along the national one.
Australia has exactly two qualifications that lead to an unrestricted electrical licence. UEE30820, Certificate III in Electrotechnology Electrician, is taught nationally. MEM31219, Certificate III in Engineering — Industrial Electrician, is taught only in Western Australia. They share not one unit code.
Every MEM31219 unit was scored against every UEE30820 unit — 3,737 comparisons. Each MEM unit keeps only its single best match. What follows is those 37 units, measured three different ways.
Each dot is a single unit of competency in MEM31219 — 28 core units every graduate must complete, and 9 electives from the industrial electrician stream.
Nothing is measured yet. This is just the curriculum — the things a Western Australian apprentice has to prove they can do.
Now each unit slides to its best composite match against the national standard. The scores compress into a narrow band: 53% to 71%, with nothing above and nothing below.
The industry convention treats 70% as the threshold for substantial similarity. Only two of 37 units reach it — terminating and testing electrical wiring, and selecting cable types and sizes. Both are core. The other 35 sit in partial overlap.
Swap the measure. Ask the language model only whether two units are about the same thing, and the picture inverts: every dot swings to the right.
Average semantic similarity across all 37 units is 87%, and not one falls below 74%. On content, these two qualifications are unmistakably covering the same electrical territory.
Now ask a narrower question: do the two qualifications test that knowledge the same way — the same steps, in the same order, marked off the same list? Performance-criteria alignment averages 6%. Five units score exactly zero — but those five are legacy-format units whose criteria the parser could not read at all. Drop them and the remaining 32 still average only 7%.
Same territory, different maps. MEM speaks manufacturing — fault-find, rectify, work safely in manufacturing. UEE speaks electrotechnology — verify compliance, AS/NZS standards, low voltage circuits. The licensing framework treats the two vocabularies as interchangeable.
Weight those signals and the 37 units resolve into a single number: 61% average alignment with the national standard.
The remaining 39% is competency with no meaningful national equivalent. The framework accepts that gap and issues a licence to work live and unsupervised on high-voltage systems.
The two measures disagree because they are asking different questions. Units that mean the same thing but share almost none of the same assessment are what you get when the same knowledge is written down twice, by two different committees, in two different house styles. It is not what you get from two different trades.
That is a defensible reason to license both. It is also, in numbers, an admission that the bar for a licence has never been unit-for-unit equivalence.
A Certificate II in Electrotechnology scores 79% against the national Cert III standard — 18 percentage points above MEM31219. It is a lower qualification, and a shorter one, and nobody suggests it should confer the same privileges. But it has no restricted pathway either.
The data does not say this is wrong. It asks a narrower question: if alignment with the national standard is not the criterion, what is?
An average tells you how big a gap is, not where it sits. Trace each of MEM31219’s 28 core units to the national unit it matched best, and you can see where.
MEM31219’s 28 core units on the left; the UEE30820 units they matched to on the right. Every unit gets a ribbon of the same width; only the colour changes — gold where the match is substantial, grey where it is partial.
Nothing here is orphaned. Every WA core unit has something on the national side that resembles it.
But the mapping is not one-to-one. Twenty-eight units land on only 22 distinct national units, because several MEM units converge on the same UEE target.
This is what the mismatch looks like close up. Where WA splits a skill across several units and the national qualification keeps it as one, the algorithm has nowhere to put the leftovers.
The clearest case. Apply safety practices for licensed electrical work, Apply OHS regulations, Work safely in manufacturing, and Plan work activities — four separate MEM core units — all match best to a single national unit: UEECD0007, apply work health and safety regulations.
Three more — Interact with computing technology, Perform computations, Organise and communicate information — collapse onto UEECS0033, use engineering applications software.
This cuts both ways, and it is worth being careful about which way you read it. It can mean WA teaches a competency in finer grain than the national standard does — four distinct safety units where UEE30820 has one. It can equally mean the algorithm ran out of plausible matches and defaulted to the nearest generic unit.
Both readings land in the same place: the two qualifications cut electrical work into different pieces. Calling them equivalent is a judgement somebody has to make and defend. The arithmetic will not make it for them.
If 61% alignment is sufficient for an unrestricted electrical licence, the obvious question is who else clears it. Using the identical method, every Certificate III in the national system that shares units with either qualification was scored against all 101 UEE30820 units — 87 qualifications, roughly 700,000 unit pairs.
Every Certificate III scored against the national electrical standard, sorted into three lanes: the UEE electrotechnology family, the MEM engineering family, and everything else — meat processing, outdoor leadership, telecommunications.
All 87 clear 50%. The lowest, Certificate III in Outdoor Leadership, scores 52%. That alone should tell you what a raw similarity score is worth without context.
Now place MEM31219 at 61% — the score that currently earns an unrestricted licence — and draw the line.
It does not fall at the top of the field. Sixty-eight of the 87 qualifications sit below it; nineteen sit at or above. The WA pathway is comfortably above average, and comfortably short of the ceiling.
This is where the argument needs care. Of the 19 qualifications at or above 61%, 18 are UEE or UET — electrotechnology and electricity supply industry qualifications. Electrical Fitting at 92%, Switchgear and Controlgear at 88%, Electrical Machine Repair at 87%.
Exactly one sits outside that family: ICT30419, Certificate III in Telecommunications Network Build, at 61.1%.
So the finding is not that a meat processing certificate has a claim on an electrical licence. It is that a tight group of electrical and utilities qualifications already clears the bar the system has set in practice — and not one of them has a pathway.
MEM31219 was never the anomaly. It is one member of a cluster of electrical and utilities qualifications sitting between 61% and 92% alignment with the national standard, of which exactly one — the Western Australian one — leads to a licence.
That is not a question about one qualification. It is a question about what the whole licensing category rests on.
The full working. Every MEM31219 unit, the single UEE30820 unit it matched best, and the score behind it. Click any row for the signal-by-signal breakdown.
| Unit▼ | Title▼ | Role▼ | Score▼ |
|---|
An earlier version of this analysis weighted performance-criteria matching at 50% of the composite score, using Hungarian-algorithm assignment. That method produced a well-publicised artefact: eleven MEM31219 units returned “Provide cardiopulmonary resuscitation” as their closest national match, because sparse performance-criteria text drove those scores toward zero and a short, generic first-aid unit became the residual best fit.
Those matches were an artefact of the weighting, not a finding about the curriculum. Re-scored with embedding similarity carrying 65% of the weight, all 37 units return a plausible electrical match and the distribution tightens to 0.532–0.710. The qualification-level average moved from 0.62 to 0.61.
Every figure on this page comes from the corrected run. Earlier drafts citing eleven “no-match” units, a CPR default, or a 38–39% figure derived from that method are superseded.
3,737 unit pairs — every MEM31219 unit (37) scored against every UEE30820 unit (101). Each pair received a composite similarity score; each MEM unit retained only its best UEE match.
Five signals, weighted by information value:
65% semantic embedding (Gemini language model, 3072-dim) · 15% performance criteria TF-IDF · 10% element alignment · 7% application text · 3% knowledge & evidence
Substantially similar > 70% composite score. Partial overlap 50–70%. Below 50% treated as distinct. Thresholds set before scoring began.
87 Cert III qualifications scored using the identical method (~700K unit pairs). Per-qualification alignment = mean of each unit’s best UEE30820 match. MEM31219 re-computed at 0.61 as validation control.
Six of MEM31219’s 37 units are imported from outside the MEM package: five from the superseded UEENEEE/UEENEEG series and one (MSMENV272) from manufacturing. The five UEE-series units are held in a legacy format whose performance criteria and elements could not be parsed, so they scored on embedding similarity alone and return 0% on both structural signals — an artefact of format, not a finding. Excluding them, mean performance-criteria alignment across the remaining 32 units is 7.3%. These five carry UEE-series codes but are distinct units from their UEE30820 counterparts; no unit code is shared between the two qualifications. UEE22025 was scored against UEE30820 using the same composite method.
A composite similarity score measures textual and structural correspondence between competency documents. It is not a measure of training quality, assessment rigour, supervised hours, or electrical safety, and nothing on this page should be read as a safety finding. All source data drawn from training.gov.au national training packages.