Miros

Miros tackles a gap that traditional ecommerce search leaves wide open: shoppers in visual categories often cannot describe what they are looking for. For fashion, furniture, footwear and decor, most intent is expressed visually or through behaviour rather than keywords, so text-based search and filters miss the products a shopper would actually buy. Miros is built specifically for that situation.

The platform works in four stages. First, it extracts thousands of visual and semantic attributes from each product image without manual tagging. Next, it interprets behavioural signals — scroll velocity, clicks, dwell time and ignored items — to build a real-time preference profile for every visitor. It then resolves contradictory signals into one coherent profile, and finally feeds that single intent layer into every discovery surface on the site: search, browse, recommendations and shop-the-look.

Miros positions itself as complementary to existing search infrastructure rather than a replacement, working alongside tools like Algolia or Elasticsearch. It publishes results from randomised A/B tests across several retailers — including conversion and GMV uplift, higher average order value and improved repeat purchases — and integrates through JavaScript widgets or a GraphQL API, with the company stating it can go live in around two weeks. Pricing is enterprise and available on request.