QSRA validation and analysis
Four-framework validation, Schedule Lens diagnostics and deterministic Monte Carlo cross-checks, evaluated on your reference data under agreed controls.
Evidence & status · reviewed 16 July 2026
Implemented capability. Controlled availability. Open release gates.
This page is the public truth boundary for SOMA Risk Studio V2.12.1. Green website or engineering tests do not authorise product release, client reliance or public application cutover.
Approved scope
What may be discussed.
External-engine boundary
Risk Studio works alongside your approved engine of record; SRS quantitative results are cross-check evidence.
QCRA
Supervised evaluation while formal numerical validation completes.
QSCRA
Not currently offered.
Evidence position
What is complete — and what is not.
| Area | Public status | Meaning |
|---|---|---|
| Feature implementation | Machine-evidenced | The feature catalogue is regression-checked; this does not confer release authority. |
| Client document pack | Machine-evidenced | Three tagged PDF/UA-1 companions pass structure, link, font and bounded-copy gates; HTML remains primary and named manual review is open. |
| Private hosted candidate | Operator-restricted | A private candidate exists for controlled verification. It is not the public application. |
| Numerical assurance | Open | Independent review and acceptance of the bounded evaluation scope remain required. |
| Manual accessibility | Open | Named keyboard, screen-reader, zoom, forced-colours, motion and document review remain required. |
| Security and data residency | Open for final review | Named review must confirm the exact candidate and repository-history remediation evidence. |
| Production identity and edge | Not cut over | Public DNS/TLS and production sign-in are intentionally withheld. |
| Final GO | Not granted | A named release and client-acceptance decision is still required. |
In development
Designed, not yet built.
The same rule applies here as everywhere else: nothing is claimed until it is evidenced. No release dates are promised.
Outturn-vs-forecast calibration. Did the P80 hold? As schemes close, Risk Studio will compare realised outturn against the forecast recorded at the time — no hindsight recalculation, and every figure will carry its sample size and confidence interval. Your own schemes, your own evidence.
Workshop copilot. Live support for risk elicitation: capture three-point estimates as they are spoken, with advisory prompts when values cluster on round numbers or ranges look implausibly narrow. Advisory only — it never blocks a workshop, never edits a value, and every prompt is logged.
Anonymised benchmark library. Cross-client range benchmarks, built only from engagements with signed data-rights terms and a minimum anonymity pool. Gated until those terms exist.
Auto-drafted registers. First-draft risk registers proposed from your schedule and prior registers — structure only. The assistant never proposes a probability, impact, duration, cost or correlation; quantities stay human. Machine-drafted rows are flagged until a person accepts them.
Need the current assurance pack?
Talk to us about a first scheme. Detailed evidence is shared through the governed procurement and security process.