Recommendations

The right product, at the right moment — homepage shelves, product-page suggestions, cart cross-sells.

The five widget types#

Each widget is suited to a specific surface in your store. Pick the right type for each placement.

Similar items

Shows products that are similar to the current product (same category, similar attributes, similar price range). Works from day one — no shopper history needed.



Use on: product pages, category pages.

Recommended for you

A personalized shelf showing products this specific visitor is most likely to be interested in, based on their browsing history.



Use on: homepage.

Buy it again

Surfaces products this shopper has purchased before and is likely to repurchase. Perfect for consumables, fashion staples, recurring categories.



Use on: homepage.

Others you may like

“Shoppers who looked at this also looked at…” — context-aware recommendations given the current product or cart.



Use on: product pages, cart pages, category pages.

Frequently bought together

Products that have historically been purchased in the same order. Revenue-optimized — picks combinations that lift cart value.



Use on: product pages, cart pages.

Which widget to use where#

SurfaceBest widgetWhy
HomepageRecommended for youPersonalized first impression — your shopper sees products tailored to them
Homepage (returning customers)Buy it againRe-engage shoppers who bought before
Product pageSimilar items”Other options like this one” — comparison shopping support
Product pageFrequently bought togetherBundle the natural add-ons
Cart pageOthers you may likeAdd-on opportunities right before checkout
Category pageOthers you may likePersonalized cross-discovery

How CartAmplify learns and improves#

The five widget types come from two different families.

Content-based widgets (Similar items) compare product attributes — titles, categories, colors, materials, price ranges. They work the moment your catalog is synced; no shopper events needed.

Behavioral widgets (Recommended for you, Buy it again, Others you may like, Frequently bought together) learn from how shoppers browse and buy on your store. Until enough events accumulate (about a week of normal traffic), these widgets fall back to popularity-based or content-based results. After that, they switch to personalized predictions and continuously improve.

Activation timeline by widget typeDay 1After ~1 week of trafficOngoingSimilar itemsLive nowRecommendedFalls backOthers you may likeFalls backPersonalizedactivatesFrequently bought togetherAfter co-purchase data

What you can control#

Each widget has a small set of dashboard settings. There are no relevance sliders or scoring weights — the system tunes itself.

  • Maximum results — How many products the widget shows (typically 4–12).
  • Exclude out-of-stock — On by default; flip off for “back-in-stock” UX.
  • Exclude recently purchased — Useful on “Recommended for you” so a shopper doesn’t see something they bought yesterday.
  • Optimization goal (Growth+) — Pick what the widget should maximize: click-through rate, conversion, or revenue per impression. “Frequently bought together” is always revenue-optimized.

Setting it up#

For supported platforms (Shopify, WooCommerce, BigCommerce, Magento, OpenCart), pick the placement from a dropdown and the plugin drops the widget into your theme — no code on your side.

For custom storefronts or headless setups, your code calls the recommendations API directly. Send the widget context (widgetId + a userEvent), receive a list of product IDs, then hydrate them in your own product tiles. Tracking impressions, clicks, and conversions is done by firing the corresponding events with the returned attributionToken. See the API reference for the request shape.

Widget count limits#

TierMaximum active widgets
Starter2
Growth5
Amplify15
EnterpriseUnlimited

A “widget” here is a placement — one search bar plus one recommendation widget counts as two. You can deactivate widgets without deleting them; deactivated widgets don’t count against your limit.

Common questions#

Want the technical details?#