The economics of panels

Modular construction companies keep going bankrupt. Katerra was the loudest, but the pattern repeats. A factory gets built, the orders don't come at the rate the factory needs, the unit economics never reach the deck, the equity runs out. Modular is sold to developers as factory efficiency. It runs as expensive asset, starved of demand.

The factory is the problem. Not the technology — the asset. A panel factory is a heavy capex bet that has to run at high utilisation to clear its cost of capital. If it runs reactively — wait for orders, build to spec, ship — it never gets there. The orders arrive piecemeal, late in design, with the panel specs already drifting toward bespoke because no one upstream had a reason to commit early.

Three archetypes sit on this landscape:

• Maximally panelised. Every floor a unit module, every wall a standard size. The factory delivers, the crane stacks, the program shortens. This is the modular pitch.
• Bespoke everything. Walls drawn to geometry, floors poured in place. Architects love it. Capital doesn't, but pays anyway because the tools to do better don't exist.
• Volume builders. No factories. Efficiency comes from volume and contract structure — the same plan built five hundred times, the same crews rotating through estates, the same supplier contracts locked in for the year. Standardisation in the deal, not the geometry.

Outside tract housing, most projects are still built bespoke. The modular supplier sits downstream of a design process that doesn't decide what to panelise until it's too late to commit factory capacity. The supplier carries the asset; the developer carries the optionality. The economics never close.

The fix isn't building better factories. It's pulling the panel decision upstream — into feasibility and master planning, before the design has hardened. If a developer can commit at the planning phase that this scheme uses these floor panels, these wall panels, from this supplier, at this cost, the factory gets a demand signal eighteen months before the pour. The factory plans capacity against a real pipeline. The unit economics start to look like the deck.

This requires modelling the modular-vs-in situ tradeoff at the feasibility stage, on real geometry, with real cost. Take a corridor floor. You've already paid to tool a floor panel for the unit. The corridor sits between units. The naive answer is to pour the corridor in situ — different shape, different load condition, why complicate it. The right answer, often, is to rotate the unit floor panel 90 degrees and use it for the corridor too. The tooling is paid for. The panel is too wide? Make a narrower variant — still cheaper than a one-off pour.

But the wall thicknesses don't add up. A 2.5m floor panel against a 200mm wall gives you a 2.7m room. The next module is 5.0m, not 5.2m. Your floor panel sticks proud of the wall, or your room is undersized, or you redesign the panel — and the moment you redesign the panel, the savings disappear.

So you accept a small rebate at the facade. The wall sits inboard of the panel. You take the cosmetic hit, the facade gets a different treatment at that edge, the math is preserved. This is the kind of decision panelised construction lives or dies on, and it is almost always made in a meeting with a sketch and a feel — long after the modular supplier could have done anything with the signal.

It should be made in a model, at the planning phase. The cost of a 200mm rebate is calculable. So is the cost of a custom panel that avoids it. So is the cost of an in-situ pour that avoids both. The reason these decisions get made by feel, late, is that no one has the geometry, the cost, and the propagation rules in one environment at the stage where the modular supplier's capacity plan still has degrees of freedom.

The mechanism we use is cascading blocks. A floor panel is a block — geometry, cost, manufacturing constraint, supplier. A unit is a block that contains floor panels, wall panels, a sun shade. A building is a block that contains units. Change the floor panel definition once and the change propagates through every unit, every floor, every building. Add a rebate to a corridor wall and 192 floor panels update.

The block also carries cost, at whatever level of abstraction the deal needs:

1. Per unit.
   00k each, 96 units, $9.6M. Used when the unit is procured as a finished module.
2. Per panel. Floor panel A: $X. Wall panel B: $Y. Sun shade: $Z. Used when the unit is tendered as components.
3. Mixed. Unit-level cost for the modular part, per-panel pricing for the in-situ corridor and the custom stair.

The choice between (1), (2), and (3) isn't fixed at design time. It's a commercial decision made later, often by a different person. The model has to support all three without a rebuild.

The innovation isn't the geometry. It isn't the cost model. It's that the two live together at the planning phase, where the modular supplier still has time to do something with the signal. Tie demand to supply early enough and the factory gets volume. The factory gets volume and the unit cost actually drops. The unit cost drops and the developer's pro forma starts working. The supplier stops going bankrupt and the developer stops watching their modular partner blow up mid-project.

That is the missing piece in industrialised construction. Not factories. Not panels. The thread from feasibility model to factory capacity plan.

To see this play out concretely — designing the units, propagating panel changes through the data model, pricing at unit and panel level — the walkthrough is below.