AI-assisted merchandising
The use of artificial intelligence to decide which products to show, in what order, and to whom, without relying on manual rules for every scenario.
Definition
Definition
Think of a good salesperson in a physical store: they notice what a customer looks at, remember what they bought last time, and mentally reorder their suggestions before speaking. AI-powered merchandising does something similar for an online store, but across thousands of simultaneous visitors.
Traditional digital merchandising runs on static rules configured by hand: "when someone searches for trainers, show brand X first" or "in the coats category, push the higher-margin products to the top." These rules give you control, but they don't scale. A catalog of 50,000 products, with hundreds of possible combinations of search query, season, stock level, and user profile, generates more scenarios than any team can manage manually.
AI-powered merchandising introduces machine learning models that process real-time signals: what each visitor searches for, which products they view, what they add to their cart, what they ignore, and what users with similar patterns end up buying. Using that data, the system adjusts the ranking of search results, category pages, and recommendation carousels without anyone having to edit a rule each time. That doesn't mean human control disappears. Ecommerce teams still define the strategy: prioritize clearance stock, push a new collection, increase average margin. The AI handles the execution, product by product, user by user.
What it's used for
What it's used for
The most direct benefit is relevance at scale. A shopper searching for "moisturizer" in a cosmetics store shouldn't see the same results as someone who has made three visits comparing anti-aging products. Without AI, both see the same list. With AI, each sees a different order, aligned with their purchase intent.
In practice, this means fewer zero-result searches, lower bounce rates from category pages, and higher average order value, because complementary product recommendations are based on real co-purchase patterns rather than generic associations.
The less visible but equally concrete benefit is operational time savings. Merchandising teams stop maintaining dozens of manual rules and can focus on strategic decisions the AI can't make for them: brand positioning, campaign narrative, or selecting new suppliers.
Doofinder
Doofinder combines manual control (searchandising) with AI-driven personalization. The system collects user behavior data — pages viewed, clicks, products added to cart, and purchases — and uses it to adjust search result rankings and product recommendations in real time for each visitor.
Specific features include:
- 1:1 personalized results: Search results are reordered based on each user's behavioral profile. Two people searching for the same term see different products based on their previous interactions.
- AI-powered recommendation carousels: Product carousels on the homepage, product pages, and category pages are driven by behavioral analysis and image matching to suggest visually similar or complementary products.
- Automatic visual tagging: The AI analyzes catalog images to generate attribute tags (color, style, material), improving search accuracy and enabling merchandising to draw on metadata that would otherwise need to be entered manually.
- Searchandising: eCommerce teams can create boosting rules, pinning rules, and promotional banners within search results, and the AI complements these by optimizing the rest of the ranking.
This hybrid model — manual rules plus AI — allows the team's strategic decisions to coexist with the kind of automatic optimization that only machine learning can execute at scale.
Example
Case study
A sportswear retailer with 8,000 products launches its summer collection. The merchandising team creates a rule to ensure new products appear in prominent positions when someone searches for "running" or "trail running." That's manual merchandising.
The AI complements that rule by detecting that users who have bought trail running shoes in the last 30 days tend to be interested in technical socks and hydration vests. For those profiles, it reorders results to intersperse accessories among the shoes. For a new visitor with no purchase history, it maintains the standard order set by the team.