EGG is now assessing buildings across Sweden. See the research
Methodology · Technical brief

From raw site data to an asset-level verdict on a building's condition.

Three stages: first, converge every available source into a verified model of the building. Then simulate how each component is actually aging under its real environment — the asset-level physical risk that regional models can't see. Finally, turn that condition map into decisions: what to convert, recover and reuse — and how to price the risk.

01
Stage 1 · Perception

Teaching the system to read a building.

Goal — converge every available source on the building into one verified model, and resolve it into a labelled understanding of every structural component.

01 · INPUT Input

Every available site data

We ingest every source we can find on the building — drawings, records, imagery, surveys — and converge them into one verified BIM model that labels every column, beam and wall.

02 · INPUT Input

Verified BIM ground truth

Across past projects, precise as-built BIM models — every column, beam and wall labelled by hand — give the system its ground truth.

03 · PROCESS Process

Paired training data

Each reconstructed model is matched to its verified BIM twin, producing thousands of supervised examples.

04 · PROCESS Process

Neural network training

A model is trained on many pairs to learn how raw inputs map to real structural elements.

05 · OUTPUT Output
COL BEAM WALL

Identified components

The system reads a new scan and resolves it into concrete columns, steel beams, brick walls — element by element.

!

Working system, improving over time. Stage 1's output is not 100% precise on day one. Every new building feeds back into the training set — the model gets sharper with each project.

02
Stage 2 · Simulation

Simulating the building's actual condition.

Goal — use the identified components to simulate, as closely as possible, what shape the building is really in today.

01 · INPUT From Stage 1
COL · BEAM · WALL

Identified components

The labelled element map produced by Stage 1 — every column, beam and wall, with its material.

02 · PROCESS Process
HUMIDITY TEMP CYCLES SOIL CONTAMINATION

Environmental layers

Layer in humidity, temperature cycles, soil and ground contamination — the conditions acting on the structure.

03 · PROCESS Process

Element interactions

Define how each factor attacks each material, and how damage propagates from one component to its connected neighbours.

04 · PROCESS Process

Chain-reaction stress simulation

Build the interactions into a propagation model and run a stress simulation across the whole structure.

05 · OUTPUT Output
68%

Building condition map

A per-component verdict: what's most damaged, what's most critical, what it means for conversion feasibility.

Aim: match reality. The simulation is calibrated to track the building's actual current condition as closely as possible — measured against on-site inspections wherever available.

03
Stage 3 · Outputs

Turning condition into decisions.

Goal — once we understand a building's true condition, decide what to do with it — which buildings to convert, which materials to recover, and how to feed them back into the circular economy.

01 · INPUT From Stage 2
68%

Building condition map

The per-component condition verdict from Stage 2 — a score for every column, beam and wall in the building. It is the single starting point that every decision below is built on: which buildings to convert, which materials to recover, and how to price the risk.

02 · OUTPUT Output

Which buildings to convert

Scan a whole stock — say 100,000 buildings — and rank them by condition to surface the ~100 best candidates: the optimal balance of minimum construction cost, maximum profit, and maximum apartments, all delivered from what already stands rather than built new.

03 · OUTPUT Output

A storage of recoverable materials

Knowing each component's material and condition, we know which parts are no longer sound and can be removed — a verified inventory of materials available for new construction. The blocker has always been knowing what's available and where; this produces exactly that inventory.

04 · OUTPUT Output
CO₂

Back into the circular economy

The material inventory feeds the circular-economy sector: what can be reused, how it is reused, and how the embodied CO₂ locked into each component is recovered and kept in use — rather than re-emitted through demolition and rebuilding from scratch.

05 · OUTPUT Output

Asset-level risk profile for insurers

The Stage 2 condition map, turned into an asset-level physical-risk profile underwriters lack — exactly where catastrophe models calibrated on historical data go blind at the asset or postal-code level. It powers the work to identify and prioritize physical risk vulnerabilities, quantify potential losses — the resilience advisory 91% of insurers call their clearest opportunity.

From a single condition map, four decisions. Convert, recover, reuse — and underwrite. Each building's verdict feeds the circular economy and the insurers who price its risk.