Match the evaluation to the job’s bottleneck. The following is the desk’s recommended procurement framework, not a description of tests already undertaken by the clubs.
| Hiring signal | Buyer question | Recommended test | Useful output |
|---|---|---|---|
| Scouting leader with supplier responsibilities | What evidence must the department retain, add or replace? | Map priority recruitment tasks against current entitlements and a proposed supplier. | Scouting Subscription Overlap Audit with retain/add/replace questions. |
| Technical scout covering new markets | Can staff assess the actual target players consistently? | Review a club-approved player sample using footage, reports and available contextual evidence. | Coverage Gap Map and unresolved-evidence register. |
| Data engineer or infrastructure owner | Can this feed be joined and maintained without repeated manual intervention? | Test identities, schema definitions, updates and failure handling on authorised samples. | Player Data Joinability Check and remediation list. |
| Recruitment modeller | Does the model help on the club’s roles and competitions? | Compare supported findings with the incumbent baseline; test missingness and relevant validation boundaries. | Same-Shortlist Supplier Trial decision memo. |
| Group data leader | Can intended users and entities use the information as planned? | Map each workflow to technical access and contractual wording; escalate ambiguities to counsel. | Licence Stress Test and group entitlement schedule. |
| Supplier implementation hire | Will support make the product usable by the actual department? | Require a working integration or repeatable report during the trial, with named support ownership. | Implementation plan and acceptance checklist. |
| AI-enabled workflow remit | Are generated conclusions traceable and appropriately qualified? | Audit claim-level sources, unsupported statements and disclosure of missing evidence. | AI Scouting Report Evidence Check. |
Set pass/fail conditions before weighted scoring. For example, a trial may require a specified match sample, an acceptable identity-matching exception rate and explicit permission for the intended workflow. These thresholds should be chosen by the club; this report does not invent universal acceptable error rates.