If someone quotes you one exact number, walk away. Published case studies report a wide range, and honestly the range is the useful part.
Category explains most of it. AR that answers will-this-fit converts differently than AR that answers does-this-shade-suit-me. Implementation quality explains the rest. A janky 3D model can convert worse than good photography, and some brands measure add-to-cart while others measure conversion rate or time on page, so the numbers were never comparable to begin with.
The pattern I trust: AR lifts most where purchase anxiety is highest. Big-ticket items, fit-sensitive products, things that are painful to return. Where anxiety was low, the lift is small, because there was less friction to remove. There is also a novelty curve nobody talks about anymore. Early AR got taps just for being new. That free attention is gone, which I think is healthy. What is left earns its taps by solving real problems.
So do not build your business case on a headline stat. Ask a narrower question: does AR work for products like mine. That needs category-level numbers, not one famous case study. A mobile app to measure AR performance lets you compare AR sessions against non-AR sessions on your own products, which is the honest version of ar product visualization roi: your lift, your catalog, your margins.
ARCommerce collects the published conversion-lift figures for 68 brands across 12 industries, so you start from reported numbers instead of a vendor's slide.
Want help figuring out what AR could do for your store specifically? I consult e-commerce brands on AR implementation at arcommerce.fyi.