Polymorphic Case study
← All case studies

03 / HOUSING

Every home’s energy needs, estimated before the first panel.

We built a household energy model from scratch, grounding it in leading thermal physics research. Given what can be observed about a home—its construction, orientation, and local climate—the model estimates how much energy that home will use for heating and cooling, and how much solar it could generate. The result is that installations can be sized on evidence rather than on a guess.

The plateLeft: a steady-state heat-conduction solve over the home’s floor plan, drawing an isotherm every degree. If the envelope is leaky the simulation throws a wide warm halo; once we know the walls are insulated that halo pulls in. Right: 1,600 candidate homes drawn from the prior for this street, plotted by annual heating and cooling. Each piece of information the adviser learns pulls that cloud toward the estimate.

ClientRESIDENTIAL ENERGY PROVIDER
CapabilityENERGY MODELLING
Duration20 WEEKS
StatusDELIVERED

01 / the gap
What does this home actually need?

Most home energy systems are sized on a guess.

Households weighing solar, a battery, or a heat pump would typically get a quote based on their postcode and a recent bill. That can be badly wrong: two homes on the same street often need very different systems, and the difference only becomes clear after the system is already on the roof.

We wanted an estimate good enough to act on with minimal input data: the energy a specific home uses for heating and cooling, and what it could generate, before anyone climbs a ladder.

Fig. 01

Two homes on the same street

These two homes have the same postcode and fall in the same bill band, but one needs more than twice the heating and cooling of the other.

Annual heating + cooling

02 / the build
Research → model → calibrate → explain

Start with the physics that research already measured.

Australia’s Nationwide House Energy Rating Scheme builds on decades of thermal research across climate zones and construction types. We used this methodology as our starting point. In our model, walls, glazing, insulation, orientation, and climate surfaced as the key determinants of heating and cooling loads, while roof geometry and shading determined solar yield.

Most homes don’t come with detailed plans. So the model is designed to work from the things that can actually be observed or asked about—floor area, era of construction, roof aspect—with reported confidence estimates for every input.

  1. Stage 0101

    Research

    Thermal performance by factor type, input data compression from wider literature.

  2. Stage 0202

    Model

    Heating and cooling loads from envelope, glazing, orientation, and climate.

  3. Stage 0303

    Calibrate

    We checked the model’s estimates against real world data, and refined our tuning accordingly.

  4. Stage 0404

    Explain

    We wanted to produce an easy to digest recommendation: what to install, how much that will help, and why.

03 / the result
Installations sized on evidence rather than guesswork.

A better first recommendation for every home.

Advisers could now size systems for a particular home rather than for its postcode, and they could show each household the estimate behind the recommendation. That meant fewer systems were oversized, and homes that needed insulation before panels could be told so up front.

±12%

annual load estimate vs metered

−18%

oversized systems

69

NatHERS climate zones covered