What Parametric Design Actually Changes in a Building

What Parametric Design Actually Changes in a Building

In 1960, the Italian architect Luigi Moretti exhibited a stadium model whose form came from nineteen parameters. Two of them were the viewing angle from the seats and the amount of concrete the budget would carry. Parametric design, at its origin, was a way to let sightlines and money shape the same curve.

That is worth remembering, because the term has drifted. In most marketing material, parametric design now means a building with a fluid skin and no right angles. The method is far more useful than the look it became known for.

What parametric design is underneath the curves

A conventional drawing records a decision. The wall is here, the fin is 600 millimeters deep, the window is 1.8 meters wide. Change one of those and someone redraws everything that depended on it.

A parametric model records the relationships. Fin depth is tied to window height. Window height is tied to floor-to-floor height. Spacing is tied to orientation. Move one input and the whole facade recalculates in seconds.

Gaudí did this with hanging chains more than a century ago. Gravity computed the arches for him, and shifting one anchor point adjusted every curve in the structure at once. The computer changed the speed. The idea was already there.

In a mid-rise residential building the useful parameters are ordinary. Orientation in degrees. Floor-to-floor height. Window-to-wall ratio. Balcony depth. Fin depth and spacing. None of it requires a curve.

I came to buildings from software architecture, and the distinction is familiar. A drawing stores an answer. A parametric model stores the question.

Why the climate is the right thing to parameterize

Phnom Penh sits at about 11.5 degrees north. From roughly late April to late August the midday sun passes to the north of the city, and for the rest of the year it sits to the south. Every face of a building takes direct sun at some point in the year.

No single shading rule survives that. The north face needs one answer and the west face needs another. A design team working by hand might test three or four shading options for a whole building, usually on the elevation the client will see first.

A parametric model tied to a local climate file and an energy engine tests every face against every hour of the year. A 2026 simulation study of high-density condominiums in Bangkok, a close climate analog, ran four fixed shading types across all four orientations. With daylight held at an acceptable level, the best option on each face reduced cooling energy by between 21.8 percent on the north and 33.7 percent on the west.

The same study found something more instructive. The options that saved the most energy also left the rooms short of daylight. Its conclusion was that shading rules drawn from sun angles alone should be rechecked against a full year of daylight and energy data.

The output is a tradeoff curve

What parametric design produces first is a chart. The form follows from it.

A typical optimization run evaluates around 1,000 variants, for example 50 options per generation across 20 generations. What comes back is a short list of candidates where no objective can improve without another giving ground. Cooling load against daylight. Daylight against glare. All of it against the facade budget.

This is where the architect and the developer-investor finally read the same page. Each point on that curve has an investment level and an operating result attached. The step from a 600 millimeter fin to a 900 millimeter fin arrives with a number: kilowatt-hours per square meter per year, on a named face of the building, against a stated addition to the facade package.

For an owner holding the building, cooling load is an operating line for forty years. For a developer placing units, it shows up later in how the building resells. Some steps on the curve defend the long-horizon return and some are decoration, and the chart makes it hard to confuse the two.

The decision still belongs to people. The model only removes the excuse for making it blind.

Where the method goes wrong

A parametric model optimizes whatever it is told to optimize. The objectives come from the brief, and a vague brief produces a very efficient search for the wrong thing.

Ask the model for a striking skin and it will deliver one, often with several hundred unique panels. Each unique panel is a fabrication drawing, a spare that nobody stocks, and a replacement question at year fifteen. A facade rationalized to six panel types and a facade with six hundred can look similar in a rendering. They do not age the same way.

So the long-horizon items have to be written in as constraints before the first run. A cap on panel types. A minimum access width for a maintenance cradle. The software has no opinion about cleanability unless someone gives it one.

The same applies to the inputs. A model run on a generic weather file describes a generic city.

Timing matters as much. The study earns its keep in the first weeks of design, while massing, orientation, and floor-to-floor height are still open. Run after the structure is fixed, it can only tune details, and the large gains have already been given away.

Parametric design is a way of asking the climate a thousand questions before the building has to live with one answer.

Owners and developers evaluating a design team can ask something plain: which decisions were tested, against what, and how many alternatives were compared. The reply says more about the building in year fifteen than any rendering of its skin.

At Imajineer, the parametric study is set up from the brief, with the climate file, the maintenance constraints, and the underwriting case loaded before any form is drawn. The conversation is available when it is useful.