We Put 3D Models on 400 Product Pages. Here Is What Moved.

We rolled 3D and AR across roughly four hundred furniture SKUs over eighteen months. The business case we wrote at the start was about conversion rate. That is not where the return came from, and I think the case we should have written was about something else entirely.

Conversion moved less than we hoped

On the pages with a 3D viewer, conversion rose by a low single-digit percentage. Real, but not the step change the vendor deck promised, and easily swamped by seasonality.

Returns moved a lot

Return rate on the 3D-enabled SKUs fell by roughly a fifth. For furniture, where a return means a van, two people and a damaged item, that is worth several times the conversion gain. The mechanism was not subtle: people who placed a sofa in their own room with AR stopped ordering sofas that did not fit.

What we would do differently

• Model the high-return SKUs first, not the bestsellers. We did it backwards.
• Insist on one master model per product that serves stills, configurator and AR. Commissioning those separately tripled the cost for no benefit.
• Budget for maintenance. Materials change, and a model that no longer matches the product is worse than no model.

We eventually settled on a specialist for the pipeline rather than a generalist agency, working with AR 3D Modeling on the e-commerce catalogue, largely because they were used to delivering to a spec sheet rather than to a mood board.