What does this studio actually demonstrate?
One parametric model carrying a whole build-to-rent development, chapter by chapter. The engine generates four massing typologies inside the planning envelope and scores each for GFA, unit yield, facade area, cost index, solar access and wind exposure. The chosen form is then cut into real floorplates with a lift core sized from the units above it, two fire stairs, a 1.8 m double-loaded corridor and units placed to a mix target. Facades panelise into a unitised kit, quantities drive a 4D delivery programme, and the pro forma closes the loop. Nothing is pre-baked: change the height limit in chapter one and the yield on cost in chapter six moves.
How is yield on cost computed here?
Bottom up, from the geometry. The cost plan applies rates per square metre of GFA by element group (structure, facade, services, fitout) to the quantities the engine produced, adds preliminaries and escalation, and adds land, so total development cost is a consequence of the design rather than an input. Income is a rent roll built from the actual units cut on each plate at weekly rents by type, adjusted for the site, plus car parking and storage income, less vacancy and an operating expense ratio. Stabilised net operating income divided by total development cost gives yield on cost, and the sensitivity view swings rent, cost, programme and cap rate to show which one the return really depends on.
What does DfMA panel repetition mean for delivery?
Repetition is what makes a facade a manufactured product instead of a construction activity. The studio panelises every elevation onto a 1.5 m module at 3.3 m floor to floor, assigns each bay a family (glazed, solid insulated, balcony or corner), then counts how few unique panel types deliver the whole building. High repetition means one panel is drawn, prototyped and tested once, then produced hundreds of times at a steady factory rate while the structure is still rising. In the programme chapter that shows up directly: unitised facade and bathroom pods compress the facade and fitout durations against the stick-built comparison, and the studio reports the weeks saved and what each week is worth.
How would this deploy into a client BIM ecosystem?
As an engine, not a website. The geometry, plates, panel schedule, programme and cost model are pure functions over a single configuration object, so the same code can drive a Revit or IFC exchange, publish quantities to a cost system, push panel families to a fabricator, or feed programme data to a scheduling tool. In client work we replace the distilled benchmark assumptions with the project's own planning controls, unit standards, cost plan and supplier rates, connect it to the live model and data environment, and land the results as native model elements and reports rather than screenshots.