Caleres Leverages AI Search to Transform Product Discovery and Raise Revenue 21%

In ecommerce, a weak search bar is usually treated like a usability issue. For brands with deep catalogs, it is closer to a revenue leak. Shoppers arrive with intent, type what they want, and then hit a system that cannot interpret style, size, width, or context well enough to guide them. They bounce, narrow the wrong way, or leave for a competitor. In Coveo’s 2025 commerce relevance research, 43% of shoppers said they go straight to search when they have a goal in mind, and 72% said they will leave if search falls short. For a multi-brand retailer, that is not a design flaw. It is a merchandising problem with a direct sales cost.

Caleres had exactly that problem at scale. The St. Louis footwear company runs a large portfolio that includes Famous Footwear, Sam Edelman, Naturalizer, Allen Edmonds, Vionic, and Veronica Beard, and sells through nearly 1,000 stores, department-store partners, branded ecommerce sites, and third-party platforms. Across its digital footprint, the company was dealing with more than 600,000 SKUs organized across color, size, width, and store availability. That complexity makes traditional rules-based search brittle very quickly. A shopper looking for “black block heel ankle boots for wide feet” is not just issuing a keyword query. She is expressing multiple layers of purchase intent at once. If the system cannot parse that correctly, the brand loses the sale long before pricing or creative become relevant.

The operating change was to replace heavy manual merchandising logic with an AI-driven search and discovery layer. By May 22, 2025, Coveo said Caleres was using its AI-relevance platform to power search and experience across the next generation of its ecommerce stack, after relying previously on thousands of manual rules that merchandisers had to maintain by hand. The company’s public description of the challenge is revealing: Caleres could not easily adjust, test, or access data across onsite search, product-listing pages, and product recommendations, despite the complexity of the catalog. In practice, that meant the digital storefront was carrying a hidden operations tax. Teams were spending effort maintaining search behavior instead of improving assortment visibility and conversion.

The AI technique here was not a generic chatbot or a vague personalization layer. It was a combination of AI search, machine learning-based relevance ranking, recommendation models, and merchandising signals. Coveo frames this as AI-relevance: a system that uses behavioral and contextual data to improve product discovery and deliver more relevant search results, recommendations, and generative experiences. Retail reporting in early 2025 described Caleres’s implementation in similar terms, highlighting AI-powered product search and discovery across 13 branded ecommerce sites, with machine learning used to support more personalized product finding. The business gain came from moving away from static search logic and toward a system that continuously learns from intent, interaction patterns, and catalog structure.

That distinction matters because product discovery is not a front-end problem alone. It sits at the intersection of merchandising, catalog operations, and conversion. In a fashion or footwear business, there are too many variables for a team to optimize manually at scale. A merchandiser can create rules for some expected queries, but cannot constantly anticipate every long-tail phrase, every regional fit pattern, or every combination of color, material, style, and size preference. AI search works because it converts messy shopper language into a better-ordered product set, then keeps refining ranking and recommendations as users interact.

A realistic example makes the mechanism clearer. Imagine a shopper lands on one of Caleres’s branded sites looking for office shoes that are comfortable enough for long wear, come in a narrow width, and still feel current enough for work travel. In a rules-heavy system, the search engine may over-index on one keyword, ignore width or fit nuance, and surface a flat list of partially relevant products. In the AI-led system, the query becomes the input. The platform interprets the intent behind the phrasing, weighs product attributes and behavioral relevance, ranks results more intelligently, and then reinforces discovery through recommendations such as adjacent styles, recently browsed items, or collection-based complements. Input: an ambiguous but high-intent search. Processing: intent interpretation, relevance ranking, personalization, and recommendation. Output: a shorter path from query to credible product choice. The commercial effect is not just better UX. It is less friction between intent and purchase.

The measured results were strong enough to make the case operational rather than experimental. On May 22, 2025, Coveo said Caleres had delivered a 21% year-over-year increase in revenue and a 25% increase in conversion rate over the measured period after implementing the platform. Retail reporting separately cited a 23% increase in conversion rate and a 5.5% increase in revenue per visitor tied to the company’s AI search enhancements. The exact measurement windows are not fully detailed in the public summaries, but the direction is consistent: smarter discovery produced better commercial outcomes. For a portfolio retailer, that matters because even modest search improvements can compound across brands, traffic sources, and catalog depth.

The business logic is straightforward. Ecommerce brands spend heavily to acquire intent, then often waste it on poor discovery. If AI search raises relevance, more of that paid and organic traffic reaches product pages that actually fit the shopper’s need. That improves conversion efficiency without needing a matching increase in discounting or media spend. It also reduces the internal burden of maintaining thousands of rules by hand, which means merchandising teams can focus more on strategic curation and less on patching digital leaks. In a multi-brand environment, that shift is especially valuable because operational complexity scales faster than headcount.

There is another reason this matters for DTC operators and retail brands: search quality shapes the perception of assortment. A customer may think a site has poor selection when the real problem is poor retrieval. Caleres’s investment suggests the opposite view. Better discovery can make an existing catalog feel broader, sharper, and more useful without changing inventory depth at all. That is a meaningful lever in categories such as footwear, where fit, occasion, and style vocabulary vary widely from one shopper to the next.

The Caleres case is also useful because it shows where AI in retail becomes practical. Not in abstract brand storytelling, and not only in customer service, but in the hidden middle layer where shopper intent gets translated into commercial action. A retailer with hundreds of thousands of SKUs does not need AI because it sounds innovative. It needs AI because manual merchandising logic eventually collapses under catalog complexity. Search and recommendation are two of the clearest places where machine learning can outperform rule maintenance on both speed and scale.

That lesson travels beyond footwear. Independent brands with fast-growing assortments, DTC operators expanding into adjacent categories, and retailers trying to lift online productivity all face some version of the same problem: too much buying intent gets lost between the search box and the cart. Caleres improved that handoff by turning discovery into an adaptive system rather than a static rules engine. The result was not just a cleaner site. It was a more productive storefront.