Observed 9 September 2026. Five search queries completed, followed by inspection of one specifically identified primary PDF. Acquisition stops here for the forest review.
No PA Consulting OSDI final assessment, inception report, detailed agreed evaluation scope, or original OSDI product-evaluation framework was located in this pass. This is a search-coverage result, not evidence that evaluation did not occur. The public record found does include a later, substantive technical evaluation that expressly attributes part of its approach to OSDI. That supplies a documented methodological reuse route while leaving the original OSDI assessment gap intact.
Programme assurance and product evaluation are distinct
The [official appointment account](http[local research file] dated 23 September 2021 and authored by project delivery lead Fiona Hallett, identifies PA Consulting as OSDI's external evaluator. Its sections on the evaluator's role and working relationship describe assurance of delivery processes, resource use, spending and quality; recommendations on governance; and communication of findings to senior civil servants. DCMS says it agreed responsibilities and priorities with PA in collaboration with the Faculty/PUBLIC supplier consortium. This is a public description of the agreed relationship, not the agreement or an evaluation result. No report attachment or detailed scope instrument was linked in the inspected article body. Reading scope: entire article body, web-reader rendition; no original HTML-byte capture by this worker.
The previously retained public-gcloud-sandbox.pdf, final PDF page 13, is a separate supplier account of PUBLIC's product-evaluation sandbox. Its framework, dataset, pilot and final-report claims do not establish that PA conducted those product tests. Root inspected that original; this worker did not independently audit its full contents. Neither OSDI role is the separate CDDP commission.
A published reuse record, with evaluation limits
[Evaluation of Safety Tech Challenge Fund Round 2](http[local research file] cover dated October 2024, is a PUBLIC-authored report hosted by government. PDF page 3 says DSIT commissioned this separate, small-scale evaluation in February 2024. PDF page 10 and PDF page 33 explicitly credit OSDI practices alongside other inputs. Page 33 sets out performance, robustness, scalability, data privacy/security and user feedback criteria, with supplier surveys, interviews and quarterly reporting as evidence sources. Footnote 17 links only to the OSDI blog homepage; it does not expose the original framework.
This report contains actual assessments: PDF page 41 assigns C4FF TRL4, CameraForensics TRL5 and Vistalworks TRL5, subject to further independent testing. PDF pages 9–10 describe a late-starting, high-level evaluation and limited testing data. Page 38 labels performance figures self-reported, cautions against unsupported cross-company comparison and calls for independent verification. These are PUBLIC's bounded findings, not independently replicated scores or PA's OSDI conclusions. The visible reuse establishes an attributed methodological contribution, not adoption of an OSDI standard into binding regulation.
The acquired original is evaluation-stcf2-original.pdf: 774,350 bytes, 54 PDF pages; SHA-256 1f44c2e0c818976571274eaa5b22c3b4ceafb85b7d4c4e42e83573d2e671f0aa. One ordinary public GET succeeded. Text reading was scoped to PDF pages 1–5, 9–10, 31–34, 38, 41, 47 and 53, plus targeted web-reader findings. PDF pages 33 and 41 were rendered and visually inspected as evaluation-stcf2-pdf33.png and evaluation-stcf2-pdf41.png; the images confirm table alignment and printed-page numbering. Printed numbers are one less than physical PDF page numbers: the key approach is printed 32 / PDF 33, and the TRL table is printed 40 / PDF 41. The whole report was not comprehensively read. Embedded annotation inspection confirmed the exact footnote-17 URL http[local research file].
Search coverage and discriminating next record
Queries used: "Online Safety Data Initiative" "PA Consulting" evaluation report; "Online Safety Data Initiative" "evaluation framework"; site:gov.uk "Online Safety Data Initiative" evaluation results; "OSDI" "evaluation" "PA Consulting" report; and "Online Safety Data Initiative" "evaluation" "final" report. These returned the evaluator announcement, programme planning/longlist material, the retained supplier service definition, sector/data reviews and the STCF2 report. Search-result references to other work do not establish that work was read. Root owns the newer Data Landscape Review; no parallel reread or new branch was started here.
The next useful record family is precise: PA's OSDI inception/evaluation plan containing the jointly agreed responsibilities and priorities; its final or periodic assurance report to the responsible senior civil servants; and the departmental acceptance or management response. The expected government custodian is the department holding the former DCMS online-safety technology programme records, with PA as the evaluator counterpart. For the product strand, the original framework version, pilot-testing protocol/results and final recommendations described by PUBLIC would distinguish a commissioned prototype from demonstrated effectiveness. None was requested or contacted, and no exact undiscovered filename is invented.
The breakpoint is therefore an identifiable assurance-output gap plus a directly attributed reuse example. Do not replace the scoped gap with “no evaluation,” merge the three programmes, equate supplier self-report with replicated performance, or infer that the later STCF2 criteria exactly reproduce OSDI's missing original instrument.