How to run assessment validation that actually improves the assessment

Validation is the part of an RTO's quality system most likely to be performed properly and evidenced badly. A panel meets, works through a sample, reaches sensible conclusions, and produces a set of minutes that nobody reads again. Twelve months later the same weaknesses appear in the same tools, and the register cannot say why.

Standard 1.5 of the Standards for RTOs 2025 is the only Standard in the instrument containing the word validation. It requires the assessment system to be quality assured by appropriately skilled and credentialled persons through a regular process of validating assessment practices and judgements. Three obligations beneath it shape how the work has to be organised.

What Standard 1.5 actually requires

• 1.5(b) — every training product, at least once every five years. Coverage is the floor, not the goal. Five years is the maximum interval, not the recommended one.
• 1.5(c) — a risk-based approach, including sample size. The sample has to follow from something you can state. "We looked at a few" is not a method.
• 1.5(f) — independence in the validation determination. The person who made the judgement cannot be the person who confirms it.

Who may validate is set separately, by Section 3 of the Credential Policy — clause 3A for non-TAE training products and clause 3B for TAE. This is a different instrument from the Outcome Standards and is cited separately.

1 · Choose the sample by a rule you can say out loud

A risk-based approach means the products getting more attention are the ones where more could go wrong, and that you can explain the ordering afterwards. The factors that reasonably drive it include how recently a product was added to scope, how much it is delivered and to how many students, its exposure to a training package change, whether recognition of prior learning is offered against it, whether a safety or industry regulator sits behind the occupation, and what the enrolment data shows.

Score them, keep the rationale for each score, and version the result. The point is not the number. The point is that when someone asks why this product was sampled more heavily than that one, there is a repeatable answer with a date on it.

Sampling itself is counted in units, and a realistic panel plans for roughly four Rules-of-Evidence judgements per sampled student. That arithmetic is what turns "we will validate the qualification" into a session with a knowable length.

2 · Check eligibility before the panel, not after

The most avoidable validation finding is a panel member who should not have been on the panel. Before someone is assigned, four things need to be true, and each is checkable in advance:

• They hold the vocational competency for the product being validated.
• They have current industry skills and knowledge relevant to it.
• They meet the Credential Policy requirements at clause 3A or 3B, whichever applies.
• They are independent of the assessment judgements under review — which means knowing whether they delivered or assessed the unit themselves.

Where a warning is overridden for a defensible reason, record the acceptance against a person. An override with a name and a reason is a decision. An override with neither is a gap.

3 · Validate practices as well as judgements

Standard 1.5 names both, and panels routinely do only the second. Judgements ask whether the evidence supported the decision that was made. Practices ask whether the tool, the conditions and the instructions could reliably produce a sound decision at all. A tool that is confusing produces defensible-looking judgements that are wrong in the same direction every time — which a judgements-only review will not surface, because each individual decision matches its evidence.

Anchor both against Standard 1.4, where the principles of assessment and the rules of evidence live: fairness, flexibility, validity and reliability at PI 1.4(a); validity, sufficiency, authenticity and currency at PI 1.4(b).

4 · End with decisions, not observations

This is where most validation stops being useful. "The assessor guide could be clearer" is an observation. It has no owner, no date, and no way of being wrong later.

Every finding worth recording should leave the session as a decision with four properties: what specifically will change, who owns it, when it is due, and what evidence will show it happened. Recommendations that do not survive that conversion are the ones that reappear at the next validation.

Then connect them. A validation recommendation and a continuous improvement entry are the same object at different moments, and keeping them in separate documents is what lets the first quietly fail to become the second. Standard 4.4's systematic monitoring and evaluation is precisely the obligation to notice that.

5 · Make the next cycle easier than this one

At the end of each session, record two things beyond the findings: how long the sample actually took against the estimate, and which of the previous cycle's decisions had genuinely been implemented. The first makes your planning honest. The second tells you whether validation is improving assessment or merely observing it — which is the difference the title of this guide is about.

Qualticks does this work as a system: validation planned on a risk basis, validator eligibility enforced before someone joins a panel, and every recommendation carried into the continuous improvement register rather than left in the minutes. See assessment validation.