Five models start from an anchor — the product on the page, or the cart. They read the product, not the shopper, so they work fully for every shopper, with or without tracking consent.
Because these models need an anchor, they belong on product pages and cart surfaces. On pages without an anchor they cannot run, and the widget serves a fallback model. You can override the product-page models with your own picks per anchor product — see Hand-picked & custom models.
Similar Products
Shows like-for-like alternatives to the product on the page. The model compares the meaning of titles and descriptions, and — when visual matching is on for your store — the product images too. The shopper who is not sure about this product sees the closest other options.
The signal behind it | Content similarity between products: text meaning, plus images when visual matching is on. No order history needed. |
|---|---|
When to use it | Keep the shopper moving when the anchor product is not quite right — wrong price, wrong color, out of stock. |
Good placements |
|
Settings | Maximum results per product — 4 to 20. A cap: it can lower the number of products a widget shows, never raise it. No cap applies until you set one. See Model settings that matter. |
Data it needs | 20 products. |
Complete the Look
Shows partners from other product types: sleeping bags, sleeping pads, and lanterns for a tent. It works even before your store has order history.
The model follows your complement map: for each product type, the types that go with it. The AI suggests the map from your product types. You can change any row on the Complements tab of the Recommendations screen in the PersonalizerAI dashboard. Within each complement type, the model picks the products closest to the anchor product. The app recomputes the results every night.
On the cart drawer, the cart page, and checkout, the model reads the whole cart. There, the Cart upsells tab can set different types per product type. A type with no cart upsell row uses its Complete the Look list.
The signal behind it | Your complement map (product type to complementary types), then the closest products within each complementary type. Without a map, it pairs related products of different types. |
|---|---|
When to use it | Outfit-style catalogs: fashion, home, hobby. The cross-category partner the shopper did not think to search for. |
Good placements |
|
Settings | No on/off switch. Edit the map on the Complements tab, and the cart version on the Cart upsells tab. |
Data it needs | 30 products, across more than one product type. |
Frequently Bought Together
Shows products that real orders combine with the anchor product. The pairs come from your own multi-item order history — no guessing. This model also powers the bundle widget with its checkboxes and add-all button. See Widget catalog.
The signal behind it | Products that appear together in your store's multi-item orders, ranked by how often. |
|---|---|
When to use it | The classic basket builder below the add-to-cart button. Strongest once your order history is deep. |
Good placements |
|
Settings | Variety — moves look-alike products further down the list. It changes the order, not which products show. See Model settings that matter. |
Data it needs | 20 products and 100 orders in the last 30 days. |
Frequently Bought Together powering the bundle widget on a product page.
Complete Your Cart
Reads the whole cart — not one product — and suggests the next best item to add. A cart with a tent and a sleeping bag gets a camping stove, not another tent. Built for the cart drawer, where it is the default model.
The signal behind it | Purchase patterns scored against every item in the cart at once. |
|---|---|
When to use it | Raise order value at the moment the shopper reviews the cart. |
Good placements |
|
Settings |
|
Data it needs | 50 products and 200 orders in the last 30 days. It starts recommending once 200 different customers have bought from you in the last 180 days. A customer who orders twice counts once. |
Premium Picks
Shows higher-priced alternatives in the same category as the anchor product. The model filters the candidates for quality: strong reviews, healthy sales pace, and a similar look. The shopper who can spend more sees a reason to.
The signal behind it | Category and price position, combined with review strength, sales pace, and visual similarity. |
|---|---|
When to use it | Catalogs with clear good-better-best ranges. Lifts average order value on the product page. |
Good placements |
|
Settings | None. On/off only. |
Data it needs | 50 products. |
Related pages
- Hand-picked & custom models — override these models with your own picks per product.
- Widget catalog — the bundle widget, cart-drawer widget, and layouts.
- Model settings that matter — where to change the settings named above.