Why is recommendation AI even harder — and more essential — for secondhand marketplaces? Because nearly every listing is one-of-a-kind, so there’s no clean catalog to lean on. Vinted, Europe’s leading secondhand fashion platform, built recommendation AI for exactly that challenge as it scaled.
The result: Vinted’s AI recommendation engine processes millions of unique secondhand listings to surface relevant items, helping the platform grow consolidated revenue to €813.4 million in 2024 — up 36% (from €596.3M in 2023).
Recommending unique items#
A conventional store recommends from a structured catalog: many units of the same SKU, clean categories, predictable attributes. Vinted has the opposite — millions of unique, user-listed items, each described differently, most existing as a single unit. Recommending in that environment is genuinely hard: “more like this” can’t mean “more of this exact product,” because there isn’t any. The engine has to infer style, fit, and taste from messy, individual listings and match them to each shopper.
Getting that right is what makes Vinted’s vast inventory navigable — and it’s central to revenue that grew 36% to over €813 million in 2024.
How recommendation AI works#
Recommendation AI predicts what each shopper wants and surfaces it. For unique inventory, it must infer taste from individual listings rather than rely on catalog structure.
Three mechanics drive Vinted’s approach. The engine infers attributes from unstructured listings — style, brand, condition, fit — to understand each item. It matches taste to inventory, connecting each shopper with the unique items most likely to suit them. And it surfaces serendipity, presenting one-of-a-kind finds shoppers couldn’t have searched for by name.
Why unique inventory rewards strong recommendations most#
The strategic insight is that the harder the catalog is to search, the more recommendations matter. On Vinted, a shopper can’t search for a specific product that exists only once — so proactive, taste-based recommendation is the primary way they discover. A strong engine turns an unsearchable sea of unique items into a personalized, navigable feed of finds. That’s why recommendation AI is existential for secondhand marketplaces, not just helpful.
What this means for your store#
If your inventory is varied, unique, or hard to categorize, this applies directly:
- Infer taste and attributes from unstructured or varied listings, not just clean catalog data.
- Match each shopper to the items most likely to suit them, even one-of-a-kind ones.
- Surface serendipitous finds shoppers couldn’t have searched for by name.
The harder your catalog is to search, the more recommendation AI matters. It turns unique inventory into discovery.
Bring recommendation AI to your store with CartAmplify#
CartAmplify brings recommendation AI that handles unique and varied inventory to any store — Shopify, dropshipping, or marketplace. Taste-based discovery that surfaces the right finds, the way Vinted navigates millions of unique listings.
Related reading#
- How Vinted’s AI Browsing Surfaces Millions of Listings
- How Sephora Drives 3.2x Purchase Likelihood with AI
Revenue figures cited from publicly reported Vinted 2024 results (€813.4M, +36%; €596.3M in 2023). Results vary by catalog, traffic, and implementation.
