Trust · The twin itself
A twin is not a data lake.
Every Digital Twin carries the record of its own making — where each layer came from, under what consent it may be used, which models touched it, and how certain any of it is.
01
Data sourcesWhich layers are populated, and which are absent — absence stated rather than imputed.
02
ProvenanceOrigin, method, instrument, pipeline version and quality flags for every observation.
03
Consent statusWhat is shared, with whom, for what purpose and until when — evaluated on every access.
04
Twin versionWhich state of the twin a given output was produced against.
05
Model versionsName, version and weights hash for every inferential component involved.
06
ConfidenceCalibrated intervals on every prediction, not point estimates presented as certainty.
07
UncertaintySeparated into uncertainty from missing data and uncertainty inherent to the model, because the responses differ.
08
PredictionsWhat was forecast, when, against which horizon, and what actually happened.
09
Simulation boundariesWhere the model declines to predict, and the specific reason it declined.
10
Human oversightWho reviewed, who edited, who rejected, and who signed.
11
Audit historyEvery access and assertion, recorded immutably and available to the person and the custodian.
Meaningful control
Four questions, answered by the individual.
What data
Chosen layer by layer, not as a single switch.
Who can use it
Which institutions, which roles, which named individuals.
For what purpose
Care, research or population analysis — separately granted.
For how long
With expiry, and revocation that propagates to derived artefacts.
Portability
Your complete twin is exportable in open standard formats on request. No customer and no individual surrenders portability as a condition of use.
Get started
Read the governance in full.
Consent architecture, provenance model and control mappings.