Skip to content

What is computational design? Parametric, generative and algorithmic design explained

Computational design, parametric, generative and algorithmic design explained: how they relate, how they compare with traditional CAD, and what they bring from design to fabrication and detail.

In short

Computational design is designing with computing power: you describe the rules and relationships of a design, and the computer builds, varies and checks it for you. It is the umbrella term for three ways of working that partly overlap.

  • Parametric design: the design depends on parameters (sizes, counts, angles). Change one and the whole model updates.
  • Generative design: the computer creates many variants from rules and goals, and helps you choose the best.
  • Algorithmic design: the design is written out as a series of steps (an algorithm), in code or in Grasshopper. That is where parametric and generative design overlap.

Updated on 19 November 2026

Working on a project where this comes up?

Whether it is a parametric model, a generative study, fabrication optimisation or smarter detailing: tell us in a few sentences what you want to achieve. You'll get a reply with a first approach within two working days, without obligation.

How the terms relate

Diagram: computational design is the umbrella term. Within it, parametric design and generative design overlap; the overlap is algorithmic design.Computational designParametricdesignGenerativedesignAlgorithmicdesign
Computational design covers parametric and generative design. Where the two overlap and the design is written out as an algorithm, we speak of algorithmic design. After the classification by Caetano, Santos and Leitão (2020).

What is computational design?

In computational design you use the computer not just to draw, but to compute the design. A CAD drawing is handwork on a screen: you place every line yourself. In computational design you define how the design comes about: which dimensions drive it, which rules apply and how parts depend on each other. The computer carries that out, and can do it again for every variant.

That brings three things: speed (a variant takes seconds instead of hours), control over complexity (double-curved shapes, thousands of unique parts) and reliability (drawings, dimensions and production files come from one source). Computational design is used in architecture, construction, shipbuilding and product development.

Parametric design

In parametric design you build a model that depends on parameters: input values such as a length, a count, an angle or a material thickness. Everything in the model is linked to them. Make a facade wider and the mullions shift, the panels adapt and the parts list recalculates.

The big advantage: you don’t design one solution but a family of solutions. That is why parametric design is the basis of product configurators, where a customer picks the sizes and instantly gets a correct model. See for example parametric models for product configurators.

In Rhino3D, parametric design is usually done in Grasshopper; in construction also with Revit/Dynamo or a live link to Tekla Structures.

Indirect manipulation: the attractor example

In traditional modelling you change elements directly: select, move, scale. In parametric design you often work indirectly: you change a driver, and the model works out what that means for every part.

The classic example is the attractor: a point or curve that influences elements based on distance. Take a facade with hundreds of panels. Close to the attractor point the openings become small, further away larger, for example for more shade on the sunny corner and more view where you want it. Move the point and all panels update at once, always consistently. By hand you would have to adjust every panel separately, and again with every change.

The same principle works with other drivers: a curve that sets the height of a building, a sun study that rotates louvres, forces in a structure that determine the thickness of members, or the curvature of a hull that determines the panel layout of a cover. The designer steers the intent; the model takes care of the execution.

Grid of circles whose size depends on the distance to one attractor point: small close by, larger further away.Attractor
Every circle responds to its distance from one attractor point. Move the point and the whole pattern follows: you steer hundreds of elements with one action.

Generative design

In generative design the computer goes one step further: it comes up with variants itself. You set rules, constraints and goals, for example as little material as possible, as much daylight as possible or as few different parts as possible. The software generates hundreds or thousands of solutions, evaluates them and shows the best. The designer chooses, adjusts and runs it again.

Generative design often uses optimisation (for example evolutionary solvers such as Galapagos or Wallacei in Grasshopper) and sometimes AI. An example from our own work: generative assembly, a database-driven approach to zero-waste design, where the design is built from available leftover materials.

Algorithmic design

In algorithmic design the design is written out as an algorithm: a series of steps the computer carries out. That can be code (Python, C#) or visual, as a Grasshopper definition. The algorithm is the design; the geometry is its result.

That is why algorithmic design sits where the other two overlap. An algorithm has inputs, so it is parametric: change the inputs and a different design comes out. And an algorithm can create variants itself, so it is generative too. Not every parametric model is a written-out algorithm (think of a CAD model with a few linked dimensions), and not every generative design is written by the designer (think of an AI tool). The overlap, where the designer states the rules explicitly, is algorithmic design.

An example: automatic generation of cut lines for a boat cover, where scripts automatically generate hems, overlaps and zip flaps for the CNC cutter from the panels of a 3D scan.

The differences at a glance

ParametricGenerativeAlgorithmic
Key questionWhat happens if I change this dimension?Which solutions are possible, and which is best?How do I describe the design as steps?
Who makes the variants?The designer, by changing parametersThe computer, from rules and goalsThe algorithm the designer writes
Typical resultOne smart model with many settingsA set of evaluated optionsA reusable definition or script
ExampleProduct configuratorOptimising a facade for daylightCutting shapes from a 3D scan

Compared with traditional CAD modelling

Computational design doesn’t replace CAD; it works on a different level. In CAD you draw the result; in computational design you describe how the result comes about.

Traditional CADComputational design
Way of workingDraw and edit elements directlyDefine rules and relationships; the model builds the geometry
A changeApplied by hand to every part and drawingAdjust one parameter; everything follows
VariantsEvery variant is a new modelVariants are settings of the same model
Getting startedQuick to startMore upfront work: the logic has to be right first
Repetition and complexityError-prone with many unique partsThe more repetition or variation, the bigger the gain
OutputDrawings and modelsAlso parts lists, CNC files and data, straight from the model

Traditional CAD remains the best choice for a one-off, simple design that won’t change. Computational design pays off as soon as a design changes often, comes in variants, has many unique parts or goes to production. In practice you combine them: in Rhino3D you draw freely where that is convenient, and Grasshopper does the repetitive work and the calculations.

Which software do you use?

For computational design, Rhino3D with Grasshopper is the most widely used combination, especially in architecture, shipbuilding and product design: visual programming, with Python and C# when it gets more complex, and thousands of plug-ins for analysis and optimisation. In construction you also see Dynamo for Revit and links to Tekla Structures; in film and games Houdini.

You don’t need to be a programmer to start. With Grasshopper you build your first parametric models within a few days. Read how in learn Grasshopper.

From design to detail: three scales

Computational design works at every scale, from the shape of a building or ship down to the last bolt. The biggest gain often lies in connecting those levels: change the design, and the production files and details follow automatically.

Design scale: architecture and ship design

At the scale of the whole it is about form, layout and performance. In architecture: massing studies, form finding, facades and the structural layout, with variants you compare directly on daylight, floor area, weight or cost. In shipbuilding and yacht building: hull forms, deck and interior layouts, appendages, covers and sails, which differ for every vessel yet are built up by the same rules. Examples: the computational design pavilion with TU Delft and the Grasshopper course for Compromiss, a specialist in composite parts for the maritime sector.

Production scale: optimising fabrication

A design also has to be buildable, ideally with as little material, time and as few different parts as possible. Computational design helps to rationalise the geometry: dividing double-curved surfaces into flat or single-curved panels, reducing the number of unique parts, unrolling parts, nesting them on sheet material and labelling every part automatically. Cutting files for CNC router, laser cutter or cutting plotter come straight from the model. Examples: automatic cut lines for boat covers and generative assembly, where the design is built from available offcuts without cutting anything.

Detail scale: connections and parts

At detail level, computational design makes every connection fit its own situation: nodes with varying angles, bolt patterns, plate thicknesses, hems and overlaps, tolerances and clearances. What would have to be drawn hundreds of times slightly differently by hand is captured once as a rule and then applied consistently and correctly everywhere. If the design changes, the details adapt. In steel construction you can link Grasshopper to Tekla Structures, so members and connections go straight into the BIM model; see the Grasshopper–Tekla Live Link workshops.

Frequently asked questions

Is parametric design the same as computational design?

No. Parametric design is part of computational design. Computational design is the umbrella term for every way of designing in which the computer computes the design: parametric, generative and algorithmic.

What does parametric design mean?

Parametric design means building a model that depends on parameters such as dimensions, counts and angles. Change a parameter and the whole model updates automatically. You design a family of solutions instead of one fixed shape.

What is an attractor in Grasshopper?

An attractor is a point or curve that influences elements in a parametric model based on distance, for example the size of openings in a facade. Move the attractor and all elements update at once.

Is generative design the same as AI?

Not necessarily. Generative design means the computer creates variants from rules and goals. That can involve AI, but usually it means rules and optimisation set up by the designer, for example with evolutionary solvers in Grasshopper.

Do I need to be able to program for computational design?

No. With visual programming in Grasshopper you can start without code. Programming in Python or C# becomes useful for more complex tasks, and that can be learned too.

What is the difference with BIM?

BIM is about information in a building model: what each element is, what properties it has and how it is coordinated with other disciplines. Computational design is about how you generate the geometry and the decisions. They complement each other: a parametric model can create BIM objects, for example in Revit or Tekla Structures.

When is plain CAD modelling better?

For a one-off, simple design that won't change, direct modelling in CAD is often faster. Computational design pays off with frequent changes, variants, unique parts or when the model goes straight to production.

Getting started with computational design?

StudioFG gives tailored Grasshopper training and builds parametric models, scripts and integrations for your project. Tell us what you are working on and you'll get a proposal, without obligation.

A project or training in mind?

Tell us briefly what you want to achieve. You get a reply with a first approach within two working days.