Miros is a "Wordless Search" product discovery technology for e-commerce retail, primarily solving the problem of products being difficult to discover when shoppers cannot accurately describe what they want using keywords, filters, or chatbots. According to the official statement, in visual categories such as fashion, furniture, and decor, a large share of shopping intent is visual, while text can cover only about 14% of it; many shoppers never use the search box, and many search terms are too vague to return the right products.
In terms of main capabilities, Miros uses AI to extract thousands of visual and semantic attributes from product images without manual tagging; it also interprets shopper behavioral signals (such as scroll speed, clicks, dwell time, and ignored products) to decode intent in real time; it synthesizes contradictory signals into a coherent preference profile; and ultimately the same intent profile simultaneously drives all discovery interfaces, including search, browsing, recommendations, and "shop-the-look." According to the official statement, the technology can infer intent from behavior rather than prompts and works in parallel with existing search systems such as Algolia and Elasticsearch.
In terms of usage, Miros provides JS components or a GraphQL API, supports headless and traditional e-commerce architectures, and the official statement says it can go live within 2 weeks, with engineering effort of less than 1 person-week, and that a working demo can be built based on a user's product catalog before commitment. The official statement also says it can already handle 500M+ signals/month and 1B+ products, with response times of 50-300ms.
Applicable scenarios include: highly visual retail categories such as fashion, furniture, and home decor, as well as online retailers with large numbers of SKUs. The official statement presents multiple performance data points: validated in randomized A/B split tests with 6 retailers (each test with 140K-2M sessions); GMV increased by 4.5% in a mobile shopping scenario; "Buy Now" clicks were more than 31% higher than Carousell's existing similar-product algorithm; and in one test, conversion rate was higher than its on-site search.
In terms of limitations, the official materials do not provide specific pricing, free trial policy, detailed technical requirements for deployment, supported languages or regions, or data privacy and compliance explanations, nor do they explain applicability to non-visual categories or non-e-commerce scenarios. The above information is not currently provided and should be subject to official channels.
