When a 3D website is worth it, and when it is decoration

We build interactive 3D sites and we turn down requests for them regularly. Not out of modesty: because a 3D site that has no reason to exist performs worse than the flat page it replaced, and we would rather say so before the invoice than after.

The test

Most 3D on marketing sites fails this test. A rotating abstract shape behind a headline communicates nothing the headline did not already say. It costs payload, battery, a class of accessibility problems, and a slice of your engineering budget forever.

Four cases where it genuinely earns its place

1. The thing has a shape that matters

Physical products where the form is the value: furniture, hardware, instruments, footwear. A photograph is one opinion about an object. Being able to turn it answers the question a photograph leaves open, and it removes a real reason people do not buy.

2. There is a configuration to explore

Anything with variants that combine: finishes, layouts, sizes, modules. A grid of photographs cannot show a combination that has never been photographed. This is the strongest commercial case for 3D on the web and it has been for a decade, because the alternative is a customer guessing and then returning the product.

3. The subject is a space

Property, venues, exhibitions, anywhere the question is what it is like to be somewhere. Flat media is genuinely bad at conveying spatial relationships, and people know it, which is why they insist on visiting in person.

4. You are selling the capability itself

This is our case, and we should be transparent about it. Our homepage runs a continuous 3D world because we make interactive things and a claim about interactivity is worth less than a demonstration of it. If you sell 3D, or games, or motion work, the site is a portfolio piece rather than a decoration.

That case does not transfer. If you sell accountancy software, a 3D world does not prove anything about your accountancy software.

What it actually costs

The build cost is the part people plan for. The running costs are the part that surprises them.

• 
  Everyone pays, every visit. Payload and GPU time are charged to the visitor, including the ones on a four-year-old Android on transit wifi. Your own machine is the least representative device you own.
• 
  Battery and heat. A continuously rendering canvas is a continuously running GPU. On a phone this is felt within a minute, and it is the kind of thing people do not report, they just leave.
• 
  Accessibility is now your problem. Canvas content is invisible to assistive technology, and motion is actively hostile to some users. Both are solvable, neither is free, and skipping them is a decision even when it is not made deliberately.
• 
  It is a living system. Browsers change, drivers change, dependencies rot. A 3D site is closer to software than to a brochure and needs the same maintenance posture.

The two different products

There is a version of this work built to win awards, and a version built to sell things, and conflating them is how briefs go wrong.

Awards work optimises for the first ten seconds and for being remembered by other designers. It tolerates a loading screen, rewards spectacle, and is measured by whether people talk about it. That is a real business objective if you are selling craft.

Conversion work optimises for someone finding a price and a contact form on a bad connection. It tolerates almost nothing and is measured in enquiries.

You can have both, but only by deciding which one loses when they conflict. On our own site the copy wins: the world sits behind a veil that exists to keep text legible, and every visual idea that fought the reading got cut, including several we liked.

If you are still going ahead

1. 01
   Write down what the 3D communicates that the copy cannot. One sentence. If it is hard to write, that is the answer.
2. 02
   Set the payload budget and the target device now, and hold them.
3. 03
   Design the reduced-motion and no-WebGL versions as first-class states, not as apologies.
4. 04
   Test on the worst device your analytics show you, not the best one in the studio.

Do that and 3D is a strong, defensible choice. Skip it and you have bought an expensive way to make your site slower.