Skip to content
HONESTASDecision-Evidence Operating System

For the Model Risk Lead

Every year the validation dossier is rebuilt from memory, spreadsheets and people who have since left.

The runs happened. The judgements were made, argued over, and acted on. But they were made in notebooks, in chat threads and in a version of the model that has since been replaced — so once a year a team reconstructs, from what survived, an account of work nobody can now re-run. The dossier is not the expensive part. The reconstruction is.

The run ledger a validator actually asks for

Run ledger · one external model · one validation yearSeeded and replayable
Illustrative validation run ledger: each run with its pinned model version, its seed, the digest of its inputs, the verdict of re-running it, the expert judgement captured at the time, and the part of the dossier it lands in.
RunModel versionInputsRe-runJudgement capturedLands in
back-test vs experienceexternal cat model · version pinnedseed pinned
sha256:2ae7f0…91cb
VALIDcaveats recorded by a named reviewer, at the timetesting against experience
champion vs challengerchallenger version · champion versionseed pinned
sha256:70b1cc…d43a
VALIDpromotion thresholds and the decision not to promotevalidation tooling justified
sensitivity and stability sweepexternal cat model · version pinnedseed pinned
sha256:c481d2…6f07
VALIDrationale for the tools chosen over the alternativesvalidation tooling justified
steered counterfactualexternal cat model · version pinnedseed pinned
sha256:5f9a3b…10de
VALIDunmodelled perils named, with the reasoningmodel completeness
re-run after a vendor version bumpexternal cat model · version CHANGEDseed pinned
sha256:5f9a3b…10de
MISMATCHheld: the change is a capital event, not a software updatemodel change governed

Nothing in that table is written for the dossier. It is what the runs left behind on the day they were run, because a run without a seed and a pinned version is not admissible evidence and the platform will not record it as though it were.

[O] Illustrative composition, built from the components the product ships. No customer data appears anywhere on this site; the digests above are placeholders. The seeded correlated Monte-Carlo engine, the stress-scenario twin and the model-lifecycle plane are observed at ROS/src/services/gtm-genesis/blueprint/MonteCarloEngine.ts, ROS/src/services/DigitalTwinService.ts and ROS/src/routes/gpu-ml.ts.

The mechanism, in three lines

The annual validation dossier compiles itself, out of runs that were reproducible on the day they happened — rather than being reconstructed once a year from memory and spreadsheets.

First

Evidence is produced on the day

Every run is seeded and pure, its inputs are hashed, and the expert judgement attached to it is captured by the person who made it, at the moment they made it — not recalled eleven months later.

Then

Champion and challenger are one record

The comparison, the promotion thresholds and the decision — including the decision not to promote — sit in the same governed thread, so “why is this still the champion?” has a written answer.

Finally

The dossier is compiled, not written

Per model, per year, out of runs that already exist and can still be re-run bit-for-bit. What is left for a human is the judgement a human is supposed to be signing.

What compiles, and what you still have to sign

The five standing validation expectations for an external model, against what the platform compiles automatically and what remains a human judgement.
What is asked, per external model, per yearCompiles from the ledgerStays yours to sign
Demonstrate understanding of the model, its expert judgements, parameters and datathe judgements and parameters are logged where they were madethe understanding itself
Name the unmodelled elements — secondary perils, non-modelled territories, loss amplificationthe counterfactual runs and their steering are recordedthe naming, and why the list is complete
Test against experience every year, with documented justification where it failsthe back-tests, their seeds and their re-run verdictsthe remediation plan
Justify the validation tooling chosen over the alternativesthe sensitivity, stability and benchmark runs actually performedthe argument for the choice
Govern model change: a new vendor version is a capital eventthe version pin, the mismatch detection and the change recordthe board decision

shipped(shipped)drift detection — designed, not built(designed, not built)

5standing validation expectations, per external model, per year[V]
~8 weeksthe sandbox pilot cycle a first model can be run inside[V]
10,000+iterations per correlated Monte-Carlo run — seeded, pure, re-runnable bit-for-bit[O]
0irreproducible runs admitted to a dossier: no seed, no pinned version, no row[O]

One model, one validation year, inside your own sandbox. If the ledger does not survive your own re-run, there is nothing here to buy — and that is a cheaper thing to find out in eight weeks than in an audit.

The honest limit