MLCLabs RAG Evaluation Pack
Find where your RAG system loses evidence or gives a weak answer.
Review retrieval and answer quality separately using a human-review workbook, 12 benchmark cases and an original fictional mini-corpus.
What this pack helps you do
- Record retrieval results and basic document precision and recall.
- Review answers against a transparent case rubric.
- Capture failure patterns and the evidence needed for a release review.
What’s included
- Editable XLSX workbook with Cases, Run Metadata, Decision sheets, clean working areas and separate fictional examples.
- Offline PDF and editable HTML guide, text decision record, quick-start instructions, licence terms and file manifest.
- Twelve benchmark cases, a fictional mini-corpus and a benchmark/review rubric.
Digital ZIP download · Bright visual edition 0.4 · £99 including VAT.
Includes 3 hours of MGCA assessment
30 minutes preparation review, one 90-minute remote workshop, and 60 minutes written findings and action planning. This is 180 consultant minutes in total.
Bring: One retrieval/generation workflow, corpus/version summary, up to ten redacted evaluation cases and a results summary.
Receive: An evaluation gap assessment and a prioritised set of additional test cases.
Book through sales@mgca.co.uk with your order reference. Remote delivery in English, up to three participants, by mutual agreement. Redeem once per qualifying product licence/order; no redemption expiry. Full submission limits, eligibility and booking terms are supplied in MGCA_Assessment_Service.pdf.
Implementation, live-system changes, certification and unlimited follow-up are excluded. Further services are optional, separately scoped, priced and agreed. This assessment is additional to supplied-file defect support.
Who it’s for
AI engineers and reviewers assessing retrieval and answer behavior.
Software and compatibility
Open with: Spreadsheet application and a browser.
LibreOffice native recalculation checked on 12 October 2026. Microsoft Excel desktop remains unverified.
The unchanged v0.1 guide records the earlier compatibility status. Run the worked example in your intended spreadsheet application before operational use.
No macros or external workbook links. The customer ZIP includes PDF guidance and assessment documents. A separate PDF preview is available before purchase.
Scope and limits
- The pack calculates basic document precision/recall and a transparent case rubric.
- It does not calculate NDCG, statistical confidence, model-judge scores or execute a model.
- Document recall is not answer correctness.
- Use a larger representative dataset, multiple runs and specialist review for the actual application.
Licence and support
Single purchasing organisation internal reuse, including contractors working for it only. No blank-pack resale, redistribution, sublicensing or use for other client organisations.
Client-delivery licensing requires separate agreement. Supplied-file defect support: sales@mgca.co.uk. File-defect support excludes consulting, customisation, project review, deployment support, certification, response-time or future-update entitlement. Statutory rights preserved.
Full terms are included in Terms.txt.
No platform subscription or customer system access is included.
Explore the pack before buying
Use the Preview link above to read the PDF excerpts. Gallery cards summarise the supplied contents and re-typeset selected fictional examples.
Related packs for a different task
Compare supplier evidence, not just promises. £59 including VAT. Sold separately.
Put the evidence behind an AI release decision. £129 including VAT. Sold separately.