Property / Address
UPRN-style reference, address text, postcode, coordinates
Trusted Data Demonstration Workbench
Can AI safely use this Digital Twin?
This first workbench demonstration uses a digital twin and spatial data scenario to show trusted data in action: sample data is profiled, related, given semantic structure, checked for governance controls, and turned into a 2-4 week MVP roadmap.
ScenarioDataMethodWorkbenchDeliverableDecision asset
Featured accelerator
Trusted Data Discovery for UK Property Digital Twin
This workbench starts with spatially linkable, hierarchical, heterogeneous property datasets. The purpose is to discover what reality the data represents before moving into semantic layer, graph, or AI work.
The approach combines Reality Mapping with a Trusted Data Discovery Accelerator: first clarify representation, identity, coverage, and confidence, then decide the next delivery step.
Client question: what does this data actually represent, and can we trust the representation enough to move into semantic layer, graph, digital twin, or AI work?
Load ScenarioDataset InventoryReality MappingIdentity AssessmentCoverage AssessmentSemantic ReadinessAssessment Report
Demonstration pack
A consulting accelerator needs a story, data, report, and next step.
The workbench is packaged as a demonstrable consulting asset. The client value is the journey from unclear source data to a decision-ready discovery report.
Customer journeySynthetic datasetScoring with confidenceSample assessment report15-minute demo scriptNext accelerator recommendation
Reference scenario data
UK property datasets as a trusted data discovery problem.
The first scenario does not need restricted data. It can begin with synthetic or open equivalents that preserve the real challenge: identity, spatial hierarchy, coverage, temporal validity, linkage assumptions, and semantic preparation.
EPC
rating, floor area, inspection date, assessment assumptions
Energy Observations
postcode-level consumption, meter count, year, coverage flag
Postcode / Boundary Lookup
postcode to OA, LSOA, MSOA, local authority, allocation method
Census Context
snapshot geography, household count, dwelling type mix
Spatial Features
building footprint, topographical feature, capture date, source layer
Synthetic data preview
The workbench consumes a small UK property sample data pack.
These CSVs are intentionally small. They show the trusted data discovery problem: property identity, EPC evidence, postcode hierarchy, spatial features, coverage assumptions, and linkage confidence.
properties6 rows
- Fields
- 9
- Preview
- property_ref, uprn_like, address_line
property_refuprn_likeaddress_linepostcodelatitudelongitudeproperty_typesource_versionrecord_status
epc_assessments5 rows
- Fields
- 9
- Preview
- epc_ref, uprn_like, address_text
epc_refuprn_likeaddress_textpostcodeassessment_datecurrent_ratingpotential_ratingtotal_floor_area_m2inspection_confidence
energy_observations3 rows
- Fields
- 8
- Preview
- observation_ref, postcode, observation_year
observation_refpostcodeobservation_yearmeter_countelectricity_kwh_meangas_kwh_meancoverage_flagpublication_threshold_note
postcode_lookup3 rows
- Fields
- 8
- Preview
- postcode, output_area, lsoa
postcodeoutput_arealsoamsoalocal_authoritylookup_versionallocation_methodallocation_confidence
spatial_features6 rows
- Fields
- 8
- Preview
- feature_ref, geometry_type, matched_property_ref
feature_refgeometry_typematched_property_refmatch_methodmatch_confidencecapture_datesource_layerknown_limitation
dataset_inventory5 rows
- Fields
- 7
- Preview
- dataset_name, source_role, represents
dataset_namesource_rolerepresentsspatial_leveltemporal_levelidentifier_fieldsknown_control_point
Dataset roles
Each source is treated as a representation of reality.
| Dataset | Represents | Spatial level | Temporal level | Control point |
|---|---|---|---|---|
| properties | addressable property | property | current record | missing UPRN-style identifier for one record |
| epc_assessments | property energy assessment | property | assessment date | address text mismatch and validity period |
| energy_observations | energy consumption | postcode | annual observation | aggregated data cannot be treated as property-level evidence |
| postcode_lookup | spatial hierarchy | postcode to OA/LSOA/MSOA | lookup version | best-fit allocation requires confidence note |
| spatial_features | building footprint or topographic feature | feature | current capture date | spatial matching rules need governance |
property_ref: P-001 | uprn_like: UPRN-100001 | address_line: 12 Alder Street / property_ref: P-002 | uprn_like: UPRN-100002 | address_line: 14 Alder Street
epc_ref: EPC-001 | uprn_like: UPRN-100001 | address_text: 12 Alder Street / epc_ref: EPC-002 | uprn_like: UPRN-100002 | address_text: 14 Alder St
feature_ref: SF-001 | geometry_type: building_footprint | matched_property_ref: P-001 / feature_ref: SF-002 | geometry_type: building_footprint | matched_property_ref: P-002
Sample scoring
Readiness score plus evidence confidence.
The score describes readiness. Confidence describes how strongly the finding is evidenced. This distinction comes directly from the research logic behind coverage, ground truth, assumptions, and spatial linkage.
These four scores are illustrative -- generated from the synthetic sample above, to demonstrate the scoring mechanism itself, not a real assessment outcome.
Reality clarity
Illustrative score: 3/4
Identity readiness
Illustrative score: 2/4
Coverage readiness
Illustrative score: 2/4
Linkage transparency
Illustrative score: 2/4
Real validation evidence
The same accelerator, run for real.
This workbench started with a synthetic sample deliberately -- “the first demonstration should not rely on restricted datasets.” Since then, the same Reality Mapping Assessment has been run against GBlocks: ~150,000 real UK properties across 24 open-government sources, all 8 dimensions scored against live database evidence rather than illustrative placeholders.
Result: 28/32, “Ready for Accelerator Follow-up” -- the top readiness band, and the first validation evidence this accelerator has had from a working system rather than a synthetic sample.
Reality clarity
Real score: 4/4
Identity readiness
Real score: 3/4
Coverage readiness
Real score: 3/4
Linkage transparency
Real score: 4/4
Demo Workbench
Digital Twin 2.0 Semantic Readiness
A Freedo-style sample workflow that turns digital twin data into profiling evidence, relationship discovery, semantic layer candidates, graph preview, and a consulting report.
Demo questionWhich critical assets are affected by an incident, which sensors and documents provide evidence, and which governance controls are needed before AI-assisted decisions can be trusted?
sitesReality boundary and spatial context
- Rows
- 3
- Fields
- 4
- Null rate
- 0%
- ID candidates
- site_id
site_idsite_namecitycriticality
assetsPhysical assets represented in the digital twin
- Rows
- 4
- Fields
- 5
- Null rate
- 0%
- ID candidates
- asset_id, site_id
asset_idsite_idasset_nameasset_typecondition
sensorsObservation devices attached to assets
- Rows
- 4
- Fields
- 5
- Null rate
- 0%
- ID candidates
- sensor_id, asset_id
sensor_idasset_idsensor_typeunitstatus
sensor_readingsTime-based observations and evidence signals
- Rows
- 4
- Fields
- 5
- Null rate
- 0%
- ID candidates
- reading_id, sensor_id
reading_idsensor_idreading_valuereading_timequality_flag
incidentsEvents that affect assets and decisions
- Rows
- 2
- Fields
- 5
- Null rate
- 0%
- ID candidates
- incident_id, asset_id
incident_idasset_idseverityincident_typestatus
documentsEngineering and maintenance evidence
- Rows
- 3
- Fields
- 5
- Null rate
- 0%
- ID candidates
- document_id, asset_id
document_idasset_idtitledocument_typeconfidence
Semantic discovery stress test
Realistic data changed what the prototype thought it knew.
Auto-Onto is an early personal prototype for profiling candidate semantic roles and relationships. Replacing a convenient synthetic sample with an 80-row GBlocks extract exposed six inference problems, including numeric identifiers being misclassified as measurements and silently reducing relationship discovery to zero.
Decision evidence
Can automated profiling propose useful semantic candidates?
Evidence is separated from scenario choices and interpretation before it is used.
Synthetic baseline
The convenient sample produced 98 relationship candidates, but its naming patterns matched the prototype's rules.
Realistic-data failure
The first GBlocks extract produced zero relationship candidates and exposed six distinct inference defects.
Corrected candidate set
After bounded fixes, the same extract produced 36 relationship candidates and improved unit coverage.
Meaning still needs a person
Candidate discovery does not prove business meaning, ownership, acceptance or safe ontology promotion.
Sample report output
What a client-facing finding could look like.
The workbench is designed to produce concise consulting evidence: what was reviewed, what controls need to be made explicit, and what should become a practical 2-4 week MVP.
Core spatial, asset, sensor, incident, and document data can support a first semantic readiness review, but source-to-scene lineage and object identity controls should be made explicit before AI consumption.
The first semantic layer should standardise Site, Asset, Sensor, Incident, Document, Observation, and Scene Object concepts, then map them back to physical data sources and scene representations.
Priority controls include spatial metadata, persistent object identity, matching rules, scene release governance, data product ownership, and AI-ready access controls.
Example 2-4 week MVP follow-up
- Week 1: confirm decision use case, review anonymised data samples, and agree readiness dimensions.
- Week 2: build a thin-slice semantic model, relationship graph, governance control summary, and MVP backlog.
- Weeks 3-4: validate with a focused prototype covering source-to-scene lineage, semantic reuse, and report output.
Connected to the concept network
This page in the Trusted Data Framework.
Connects the prototype workflow to the concepts and consulting assets needed for digital twin semantic readiness.
Curated DataSemantic LayerKnowledgeIntelligenceDecision
Related concepts
Consulting assets
Delivery connection
AssessDiscoverPrototypeGovern
Cross-cutting
GovernanceLineageObservability