What Merchandiser AI Is
Merchandiser AI is one of the ways a collection can be sorted. Instead of following a fixed rule, a machine-learning model estimates how much demand each product is likely to attract and ranks the collection by that estimate.
You choose it per collection. Merchandiser AI can run your large category pages while a sorting rule or a manual order runs the others.
One Model per Store
Every store gets its own model, trained only on that store's products and their performance. Nothing is shared between stores.
The model is retrained regularly as new product and performance data arrives. New arrivals, price changes and shifts in demand are picked up without anyone editing a rule.
What the Model Learns From
The model looks at each product from several sides at once:
- Demand
- Views, add-to-carts, purchases, revenue and conversion rate of the product.
- Price and discount
- The sale price and the discount rate.
- Stock
- How many units are in stock.
- Product attributes
- Brand, category, season, gender and color.
- Freshness
- How many days the product has been listed.
Product data comes from your product feed. Performance data comes from Google Analytics 4, from your store platform or from the product metrics API. That API also accepts in-store sales; when you send them, the model learns from online and in-store sales together.
What It Predicts
By default the model predicts a combined demand score in which purchases count the most, then add-to-carts, then views.
The target can be set for your store instead: purchases, revenue, add-to-carts, views or conversion rate. We choose it with you when your account is set up.
Size Availability Weighs In
A strong prediction is not enough to reach the top of the page. Each prediction is weighted by the product's size availability, which is the share of its sizes that are in stock.
A product that has sold out of most sizes therefore ranks below an equally promising product that is fully in stock.
Example
-
Product A
Position 1Predicted demand
Sizes in stock: 6 of 6
-
Product B
Position 2Predicted demand
Sizes in stock: 2 of 6
You Stay in Control
- Pins come first. A pinned product keeps its position, and the model orders the rest of the collection around it.
- Boosts add your priorities. Products can carry a boost, for example to push the current season, which is added to the model's estimate. Boosts are set up with our team.
- You can preview the result. On a collection's sort page you can see the AI order before anything changes on your store. It is applied only when you save.
- You can go back. A change your team saves is kept in the version history and can be rolled back.
When to Use AI and When to Use a Rule
Merchandiser AI suits collections where many signals matter at once and no single rule describes what sells. A sorting rule suits collections with an explicit policy, such as newest first.
The model needs performance history to learn from. A new store, or a collection whose products have had little traffic, is better served by a sorting rule until that data has built up.
Merchandiser AI is included in the Professional and Premium plans. See the plans for what each one includes.