Skip to content

Our model

Grade

Grade is a stock model for the buildings we already have. It observes every unit, learns how condition and performance change, and predicts the effect of a retrofit before it is funded.

2,015,843 London units · reconciled to the GLA London Building Stock Model

How it works

A model of what happens next.

A stock model learns the dynamics of a building: given its current state and an intervention, it predicts the next state. Grade applies this to every building at once.

A London terrace as it stands today The same terrace with the proposed measures marked The same terrace after the works The same terrace with predicted performance drawn over it

  1. 01 · State

    EPC D · solid wall · gas boiler 2009 · 142 kWh/m²/yr

    Current state of the building. From the lodged certificate, the footprint and the metered neighbourhood.

  2. 02 · Action

    EWI + loft top-up + air-source heat pump · 2027 Q2

    Fabric-first retrofit. Measures the assessor already recommended, screened against heritage rules.

  3. 03 · Future state

    EPC B · 61 kWh/m²/yr · −2.1 tCO₂e/yr · payback 11.3 yrs · 82 % confidence

    Predicted after the works. Illustrative output: every figure carries its source, date and confidence.

Foundational primitives

Foundational primitives.

Three questions the model is built to answer, in the order a programme asks them.

Answer
1,284 homes across 212 streets. 61 % solid wall, 38 % with a 2019+ certificate on the adjoining unit. Source: MHCLG EPC register, OpenStreetMap footprints, AlphaEarth change rank.

Now · Infer reality

Infer reality.

Know what exists, its condition, and what changed.

Product options

How it fits your stack.

01

Integrations Coming soon

Grade’s outputs land in the systems you already use: asset registers, reporting tools and GIS.

ExcelPower BIQGISArcGISAsset-management systemsHousing-management systems

02

API

Your models, thresholds and reporting run on Grade’s outputs. One request, one building, every figure sourced.

# predicted state after a measure set
POST /v1/predict
{ "uprn": "100023336956",
  "measures": ["ewi", "loft_topup", "ashp"],
  "from": "2027-Q2" }

# → 200
{ "epc": { "now": "D", "predicted": "B" },
  "kwh_m2_yr": { "now": 142, "predicted": 61 },
  "tco2e_yr_saved": 2.1, "confidence": 0.82,
  "sources": ["MHCLG EPC 2024-03", "GLA LBSM 2026"] }

03

Web app

Start in days. No setup, no code: open a borough, rank the stock, open a building, read the record.

The web app: a building opened through the performance lens
Try for free

Grade learns how the building stock evolves and tells you what to retrofit next.

For organisations ready to plan retrofit on evidence, not spreadsheets.

Talk to us