HRM.2:4.2 - Build the evidence argument before scoring candidates
For each consequential claim, choose evidence that could support it and explain the inferential step. A representative work sample may demonstrate a defined operation; a structured interview can investigate particular prior behavior or reasoning; a qualified test can support a specified construct or predictive claim. None is universally best. Specify the conditions, relevant comparison and decision use before knowing who benefits from the rule.
| Claim needed | Possible evidence | What must be established before use |
|---|---|---|
| Performance of a representative advisory task | Work sample with permitted tools and an explained scoring basis | The task samples consequential work, scoring concerns that work, and the support corresponds to the job |
| A contribution across a wider class of cases | Relevant history or a professionally supported assessment | The evidence can generalize to the proposed work, population and conditions |
| Eligibility or a required qualification | The applicable verified fact | The rule actually requires it and the fact concerns this person and use |
| Performance after an obtainable learning period | Appropriate evidence and a credible learning/provision arrangement | The recommendation distinguishes present capability from the conditional future contribution |
Ask a qualified assessment supplier for the validity argument for this decision, population and conditions; the consistency of measurement; relevant fairness and bias findings; limitations; permissible administration and change conditions. A vendor’s reference customer or general accuracy figure is insufficient when the inference changes. The supplier can perform technical validation; HR must obtain enough of its result to understand which use is warranted. If the result is unavailable, use an adequate alternative or leave the dependent comparison unresolved.
A work sample’s resemblance to the job alone does not establish every predictive claim. Likewise, correlation with old supervisor ratings is not automatically evidence of the new contribution if those ratings reflect obsolete tasks or unequal opportunities. Keep the intended outcome and the evidence for it explicit. Use existing HCD or OPS.22 results only within the claim each supports; a good operating result may depend on team or agent contributions not controlled by the individual.