6 November 2026International edition
Vol. I · No.
6 November 2026
AI IN FASHION
DAILY
The daily briefing on AI in the fashion business
Where fashion meets artificial intelligence.
Merchandising & Buying · Stories

What is Fast Simon and how does visual search improve merchandising for Steve Madden?

The footwear retailer deployed AI visual search across 25 stores in 2024, using image recognition to surface products faster and reshape conversion patterns.

a woman looking at a pair of shoes in a shoe store
Photo: Vlad Vasnetsov / Unsplash

KEY TAKEAWAYS Summary by the editors

  1. Fast Simon provides AI-powered visual search and merchandising tools that let shoppers upload images to find similar products across inventory.
  2. Steve Madden rolled out Fast Simon's platform across 25 stores in 2024 to streamline product discovery and improve conversion rates.
  3. Visual search uses computer vision to match image attributes such as colour, shape and texture against catalogue data in real time.
  4. Retailers cite faster time to purchase and reduced bounce rates, though basket value and return rate impacts remain less documented.

Fast Simon is an artificial intelligence merchandising platform that offers visual search, personalised product ranking and site search optimisation for fashion and footwear retailers. The company, founded in 2016 and headquartered in Texas, serves over 7,000 e-commerce sites worldwide, applying machine learning to match shopper intent with inventory and to surface products that drive conversion. In 2024, Steve Madden, the American footwear brand operating more than 200 stores globally, deployed Fast Simon's visual search technology across 25 locations to help customers find styles by uploading photos from their phones or the web.

Visual search matters now because browsing friction remains a persistent conversion barrier in fashion e-commerce. Shoppers often struggle to describe a desired style in keywords, while traditional text search returns irrelevant results when colour, silhouette or material are the defining attributes. Visual search bypasses this by letting customers upload an image and receive ranked results based on visual similarity. For merchandisers, the technology also generates behavioural data about which attributes drive clicks and purchases, feeding back into automated product ranking and personalisation engines.

This article explains how Fast Simon's platform works, examines the Steve Madden deployment, assesses the evidence for conversion and merchandising benefits, and outlines the operational and accuracy constraints that fashion executives should weigh before adoption.

What is Fast Simon and what does the platform do?

Fast Simon is a software-as-a-service merchandising suite that combines visual search, predictive search, collection and product page personalisation, and AI-driven product ranking. The core technology uses computer vision models trained to recognise fashion attributes such as colour, pattern, neckline, heel height and material from product images. When a shopper uploads a photo, the system extracts these features and matches them against the retailer's catalogue in real time, returning a ranked list of visually similar items.

Beyond visual search, Fast Simon's platform includes autocomplete and instant search that learns from past queries, filters and purchase behaviour, as well as personalised collections that reorder product grids for each visitor based on browsing history and predicted affinity. The company positions itself as a plug-in layer that sits between a retailer's existing e-commerce platform, such as Shopify, BigCommerce or Salesforce Commerce Cloud, and the front-end shopper experience. Installation typically requires embedding a JavaScript tag and mapping product feeds; no re-platforming is necessary.

Fast Simon's customer base spans direct-to-consumer brands and multi-brand retailers across fashion, footwear and home categories. The company reports processing over 10 billion search queries annually, though it has not disclosed revenue or profitability figures. Investors include Moneta VC and Flint Capital, which led funding rounds in 2020 and 2021. The platform competes with specialist visual search providers such as Syte and Visenze, as well as larger e-commerce search vendors like Algolia and Bloomreach, some of which have added visual search modules in recent years.

How does visual search technology work in fashion retail?

Visual search in fashion relies on convolutional neural networks, a class of deep learning model designed to recognise patterns in images. During training, the model ingests millions of labelled product photographs and learns to associate pixel patterns with attributes such as sleeve length, fabric texture, heel type or collar style. Once trained, the model can analyse a new image, whether a shopper's Instagram screenshot or a street-style photo, extract those same attributes, and encode the image as a numerical vector, often called an embedding.

The retailer's product catalogue is processed in the same way: each item image becomes a vector in a high-dimensional space. When a shopper uploads an image, the visual search engine calculates the distance between the query vector and all product vectors, ranking results by similarity. The closest matches appear first. Some systems also apply business rules, such as prioritising in-stock items, promoting higher-margin SKUs or filtering by the shopper's size or location, layering merchandising logic on top of pure visual similarity.

Accuracy depends on the quality of the product images, the diversity of the training data and the granularity of the attribute taxonomy. A model trained predominantly on Western fast fashion may struggle with traditional garments, bridal wear or technical outerwear. Similarly, low-resolution or heavily styled campaign imagery can confuse the matching algorithm if the catalogue photos are plain white-background shots. Retailers must therefore curate training sets and product feeds carefully, and many platforms offer continuous learning loops that refine the model as users click, purchase or reject results.

What did Steve Madden deploy and when?

Steve Madden integrated Fast Simon's visual search and AI merchandising platform across 25 of its retail stores in 2024. The deployment was part of a broader digital transformation initiative to unify online and in-store product discovery. Shoppers visiting a Steve Madden store can now use in-store kiosks or their own mobile devices to upload a photo of a shoe or accessory they have seen elsewhere, and the system returns matching or similar styles available in that location or online.

The rollout addressed a specific pain point: customers often arrived with a vague description or a saved image from social media, and sales associates spent significant time manually searching inventory or flipping through lookbooks. Visual search automated that discovery step, reducing friction and freeing staff to focus on fitting and service. Steve Madden has not published detailed performance metrics, but the company stated in public communications that the technology improved product discovery speed and contributed to higher conversion rates in pilot locations.

Steve Madden operates more than 200 company-owned and licensed stores worldwide, alongside a substantial wholesale and e-commerce business. The brand reported net sales of approximately USD 2 billion in its 2023 fiscal year, with direct-to-consumer channels, including e-commerce and owned retail, representing a growing share. The Fast Simon deployment reflects a strategic bet on AI-driven personalisation and visual tools to differentiate the in-store experience and capture demand that might otherwise flow to online pure-plays or competitors with more advanced digital interfaces.

Steve Madden store footprint with and without visual search
25 storesStores with Fast Simon visual searchDeployed in 2024
200 storesTotal Steve Madden stores worldwideCompany-owned and licensed locations

Source: Steve Madden public communications, Fast Simon case materials, 2024

How does visual search improve conversion and merchandising outcomes?

Proponents of visual search argue that it shortens the path from intent to purchase by eliminating the need to translate a visual idea into keywords. A shopper who uploads an image of a chunky platform sandal sees relevant results immediately, without trial-and-error refinement of text queries. Faster discovery theoretically reduces bounce rates and increases the likelihood that a session ends in a transaction. Some vendors cite case studies showing double-digit improvements in conversion rates, though these figures often reflect controlled tests or single-retailer anecdotes rather than industry-wide benchmarks.

For merchandisers, visual search generates a new data stream: which images shoppers upload, which attributes the algorithm identifies, and which results they click or ignore. This feedback can inform assortment planning, highlight gaps in the catalogue and reveal emerging trends. If many users upload images of square-toe mules but the retailer stocks only pointed styles, the merchandising team receives a clear signal to adjust future buys. Similarly, visual search logs can surface popular colourways or silhouettes that text search alone would miss.

Visual search also supports cross-sell and upsell strategies. Once a shopper finds a base item through image upload, the platform can recommend complementary products, complete-the-look ensembles or higher-priced alternatives with similar visual features. This layering of recommendations on top of search results mimics the associative browsing that happens in physical stores, where a customer picks up one shoe and the sales associate brings three more. The digital equivalent, powered by real-time matching and personalisation, aims to replicate that serendipitous discovery at scale.

What are the documented results and return on investment?

Published data on the return on investment from visual search remain sparse and largely vendor-supplied. Fast Simon's case materials and public statements reference improvements in key performance indicators such as conversion rate, average session duration and repeat visit frequency, but the company has not released aggregated statistics across its customer base. Steve Madden's deployment has not been accompanied by a detailed financial disclosure of incremental revenue or margin impact attributable to the visual search rollout.

In the broader market, a 2025 report by McKinsey on AI in retail noted that visual search and personalisation engines, when properly implemented, can lift conversion rates by 10 to 30 per cent in controlled tests, though the effect size varies by category, baseline experience and integration quality. The report cautioned that many early deployments suffered from poor image quality, narrow training sets or misaligned business rules, diluting the potential benefit. McKinsey recommended that retailers conduct rigorous A/B testing, segment results by device and customer cohort, and isolate visual search impact from concurrent site optimisations.

Return rate data are especially scarce. Visual search could theoretically reduce returns by helping shoppers find products that better match their intent, but it could also increase returns if the algorithm surfaces items that look similar in a photo yet differ materially in fit, fabric or construction. Retailers have not publicly correlated visual search adoption with changes in return rates, leaving this question open for now.

Which other fashion retailers are using AI visual search?

Several major fashion retailers have integrated visual search into their digital channels over the past five years, though the extent of deployment and the technology providers vary. ASOS, the British online fashion retailer, launched a visual search feature in its mobile app in 2017, allowing users to snap a photo or upload an image to find similar items across its 85,000-plus product range. The company reported that visual search users exhibited higher engagement and conversion than those relying solely on text search, though ASOS did not disclose precise figures.

Zalando, Europe's largest online fashion platform, introduced visual search in 2019 across its app and website, initially in partnership with a third-party provider and later integrating the capability into its in-house technology stack. Zalando's implementation emphasises real-time personalisation, combining visual similarity with individual browsing history and size preferences to rank results. The company has stated that visual search contributes to its broader personalisation strategy, which aims to surface the most relevant items from a catalogue exceeding 500,000 products.

In the United States, Nordstrom piloted visual search in 2018 and subsequently rolled it out across mobile and desktop. The retailer's approach integrates visual search with its existing style recommendation engine, so that a shopper who uploads a dress image also receives suggestions for shoes, bags and jewellery that complement the look. Nordstrom has not published standalone visual search metrics, but the company regularly cites AI-driven personalisation as a driver of customer satisfaction and repeat purchase rates in its investor communications.

Luxury brands have been slower to adopt visual search, in part because of concerns about brand control, image quality and the risk of surfacing third-party or counterfeit products in search results. Some high-end retailers have implemented invite-only or logged-in visual search features that restrict results to authenticated inventory, balancing convenience with exclusivity.

Selected visual search deployments in fashion retail
  1. 2017ASOS launches visual search in mobile app for 85,000-plus products
  2. 2018Nordstrom pilots visual search and integrates with style recommendation engine
  3. 2019Zalando introduces visual search across app and website in Europe
  4. 2024Steve Madden deploys Fast Simon visual search across 25 retail stores

Source: Company announcements, trade press reports, 2017-2024

What are the risks, limitations and open questions?

Visual search is not a panacea, and several technical and commercial constraints temper the enthusiasm. First, accuracy remains uneven. Models trained on generic fashion imagery may misclassify niche categories, luxury materials or culturally specific garments. A shopper uploading an image of a traditional embroidered blouse might receive results for a printed T-shirt if the training data lacked sufficient examples of that style. Retailers must invest in domain-specific training and ongoing model refinement to maintain relevance, which requires both data infrastructure and machine learning expertise.

Second, visual search introduces privacy and intellectual property considerations. When a user uploads a photo, the retailer gains access to that image and potentially metadata such as location or device type. Clear data handling policies and user consent mechanisms are essential to comply with regulations such as the General Data Protection Regulation in Europe and emerging privacy laws in the United States. Additionally, if a shopper uploads a competitor's product image or a designer's runway photo, the visual search engine may surface look-alike items, raising questions about design appropriation and brand differentiation.

Third, the business case depends on baseline conversion rates and the quality of the existing search experience. A retailer with a strong text search engine, comprehensive filters and high-quality product imagery may see only marginal gains from adding visual search. Conversely, a site with poor search and navigation can achieve dramatic improvements, but those gains may come as much from fixing foundational issues as from the novelty of image-based queries. Executives should benchmark current performance, run controlled pilots and isolate variables before committing to enterprise-wide deployment.

Fourth, visual search adoption among consumers is still maturing. While mobile-native shoppers readily embrace photo-based discovery, older demographics or those shopping from desktop may default to text. Usage rates vary widely by region, category and device, so retailers must tailor their user interface and marketing to encourage trial. Some platforms report that fewer than 5 per cent of site visitors use visual search even when prominently featured, though those who do often convert at higher rates.

Finally, integrating visual search with legacy systems can be complex. Retailers running on older e-commerce platforms or fragmented technology stacks may struggle to pass real-time inventory, pricing and personalisation signals to the visual search engine. Latency, synchronisation errors and inconsistent product data can degrade the user experience, undermining the intended benefit. A successful rollout requires close collaboration among merchandising, IT, and vendor teams, as well as rigorous testing across devices and geographies.

How does Fast Simon compare with other visual search providers?

The visual search and AI merchandising market is fragmented, with specialist vendors, enterprise commerce platforms and cloud infrastructure providers all offering solutions. Fast Simon competes primarily with Syte, a Tel Aviv-based company founded in 2015 that serves retailers including Farfetch and Marks & Spencer, and Visenze, a Singapore-headquartered provider with clients in Asia and Europe. All three focus on fashion and lifestyle categories and offer both visual search and product recommendation modules.

Broader e-commerce search platforms, including Algolia, Bloomreach and Coveo, have added visual search capabilities through partnerships or in-house development. These vendors pitch an integrated approach, combining text search, navigation, personalisation and visual discovery in a single contract. The trade-off is depth versus breadth: specialist providers typically offer more sophisticated computer vision models and fashion-specific taxonomies, while integrated platforms simplify vendor management and data flows at the cost of some feature richness.

Cloud providers Amazon Web Services, Google Cloud and Microsoft Azure offer visual search APIs and pre-trained models that retailers can embed in custom applications. This approach gives maximum control and flexibility but requires engineering resources to build and maintain the front-end experience, ranking logic and feedback loops. Smaller retailers and brands typically lack the scale and expertise to pursue this route, favouring turnkey SaaS options like Fast Simon.

Provider typeDeployment modelFashion specialisationTypical customer
Specialist visual search (e.g. Fast Simon, Syte, Visenze)SaaS plug-inHigh: fashion-trained models and attributesMid-market and enterprise fashion retailers
Integrated commerce search (e.g. Algolia, Bloomreach)Platform with modulesMedium: general retail with fashion supportMulti-category retailers and large brands
Cloud provider APIs (AWS, Google, Azure)Custom integrationLow: generic models requiring trainingEnterprises with in-house engineering teams

What operational changes does visual search require?

Implementing visual search is not solely a technology decision; it demands changes to workflows, content production and performance measurement. Merchandising teams must ensure that product images meet the visual search engine's quality standards: consistent lighting, neutral backgrounds, multiple angles and high resolution. Many fashion brands shoot aspirational campaign imagery that is striking but visually noisy, with models, props and dramatic styling that confuse the matching algorithm. Retailers often maintain separate image sets for visual search, adding cost and complexity to the content supply chain.

Data governance becomes more critical. The visual search engine requires access to real-time inventory, pricing, size availability and product metadata. Stale or incomplete feeds degrade the experience, surfacing out-of-stock items or showing incorrect prices. Retailers must establish synchronisation protocols, monitor data quality and resolve discrepancies between systems. This coordination spans merchandising, IT, warehouse management and e-commerce operations, requiring cross-functional alignment and clear ownership.

Customer service teams need training to support visual search queries. Shoppers who upload an image and receive unexpected results may contact support, and agents must understand how the technology works, what factors influence ranking and how to escalate false positives or negatives to the platform provider. Some retailers integrate visual search logs into their customer data platforms, allowing service agents to see what a shopper uploaded and tailor their response accordingly.

Performance measurement must evolve as well. Traditional e-commerce metrics such as search-to-purchase conversion, bounce rate and average order value still apply, but visual search introduces new dimensions: image upload rate, match accuracy, result click-through and repeat usage. Retailers should define success criteria before launch, establish baseline measurements and conduct regular A/B tests to isolate the impact of visual search from other site changes. Attribution can be tricky, especially when a shopper begins with text search, switches to visual search and completes a purchase after viewing recommendations, so multi-touch analytics are essential.

Frequently asked questions

What should fashion companies do now?

  1. Audit your current search performance by measuring text search conversion, bounce rate and query refinement frequency across devices. Identify pain points where shoppers struggle to describe or find products, particularly for visually distinctive categories such as footwear, accessories or patterned apparel, to establish a baseline for improvement.
  2. Evaluate specialist visual search providers by requesting live demos with your own product catalogue and conducting pilot tests with a subset of customers or geographies. Compare accuracy, latency, ease of integration and pricing models, and insist on transparent reporting of match quality and usage metrics before committing to an enterprise contract.
  3. Invest in high-quality product imagery by standardising image specifications, shooting multiple angles and ensuring consistent backgrounds and lighting. Work with your creative and content teams to produce a parallel image library optimised for algorithmic matching, separate from campaign or editorial assets, and maintain version control as styles are updated or retired.
  4. Define clear business rules and merchandising logic that layer on top of visual similarity, such as prioritising in-stock items, promoting higher-margin SKUs or filtering by customer size and location. Collaborate with your merchandising and planning teams to encode these rules in the platform configuration, and review and adjust them quarterly based on performance data and inventory position.
  5. Establish data privacy and consent protocols by drafting clear terms that explain how uploaded images are used, stored and deleted. Consult legal and compliance teams to ensure alignment with GDPR, California Consumer Privacy Act and other regulations, and implement user-friendly consent flows that do not create friction while protecting customer rights and your brand reputation.
  6. Train cross-functional teams including merchandising, IT, customer service and store associates on how visual search works, what results shoppers can expect and how to troubleshoot common issues. Create internal documentation, run hands-on workshops and designate champions in each function who can serve as first-line support and feedback channels to the platform provider.
  7. Monitor and iterate continuously by tracking visual search adoption, match accuracy, conversion lift and customer feedback. Run controlled A/B tests to measure incremental impact, segment results by customer cohort and product category, and feed insights back into model training, business rules and site design. Plan for quarterly reviews and be prepared to refine or pivot if usage or return on investment falls short of expectations.

Frequently asked questions

Does visual search work on mobile devices?

Yes, most visual search platforms are optimised for mobile, allowing shoppers to upload photos from their camera roll or capture images in real time. Mobile accounts for the majority of visual search usage because the workflow is native to phone behaviour, though desktop implementations also support drag-and-drop or file upload.

Can visual search identify products from social media screenshots?

Visual search can process screenshots from Instagram, Pinterest or other platforms, but accuracy depends on image quality, cropping and whether the original post shows the product clearly. Heavily filtered or low-resolution images may yield less precise matches, and some platforms strip metadata that could aid identification.

What happens if the visual search algorithm returns no matches?

Most platforms display a 'no results' message and prompt the shopper to try a text search, adjust filters or upload a different image. Retailers can configure fallback behaviour, such as showing the most popular items in a related category or inviting the customer to contact support. Tracking no-result rates helps identify gaps in the catalogue or model training.

Stories