Create your first campaign

Define placements, source products, candidate pools, strategy, and display settings for a recommendation rule, then preview and publish.

A campaign is the publishing container; rules inside it decide what is recommended where. A rule connects placements, source products, a candidate pool, and a strategy.

This guide covers individual Upsell recommendations; start from Create campaign → Upsell campaign. To sell a complete set at one price, re-quoted for the exact variants picked, use Bundles.

Step 1: Choose placements

Pick at least one: product page, cart page, or cart drawer. Product page rules resolve against the product being viewed; cart rules resolve against everything in the cart.

Step 2: Choose source products

The source product is what triggers recommendations—the snowboard that should suggest wax. Select individual products or start from a discovery row that already pairs a source with candidates.

Step 3: Approve the candidate pool

The candidate pool is the complete set of products the rule is allowed to recommend. Add candidates from discovery or manually. Nothing outside the pool is ever shown, and you can pin, pause, or exclude a candidate at any time without touching the rest of the rule.

Step 4: Choose a strategy

• Fixed returns candidates in your manual order. Use it when priorities are known, traffic is low, or a product must always appear.
• Rotate spreads exposure evenly across the pool. Use it to collect initial data or keep a lineup rotating.
• Optimize shifts exposure toward candidates that perform against your goal. Use it once traffic is steady and you want the pool to compete.

Step 5: Set the optimization goal

Optimize supports add-to-cart rate, purchase rate, or revenue per impression. Pick the one that matches what the placement is for—cart add-ons usually optimize add-to-cart rate; considered purchases usually optimize revenue per impression.

Step 6: Display settings

Display settings—heading, button text, item count, layout, image ratio, price and compare-at price visibility, quick add, variant selector, mobile item count, and empty state—only affect the official app blocks. API responses are unaffected.

Step 7: Filters

The storefront applies a small set of fixed exclusions; there are no per-rule filter settings to configure:

• On the product page, the product being viewed is never recommended to itself.
• A candidate with no sellable variant is dropped. If the preferred variant is sold out, another available variant is offered instead.
• A candidate Shopify returns no product data for in the shopper’s current Market is dropped.
• A product already in the shopper’s cart is never recommended, on any placement.

Optional recommendation discounts

A rule can carry a default discount. A candidate can inherit it, have no discount, or override it with a percentage or fixed amount. Review the campaign schedule and Shopify discount-combination settings alongside those rule settings. A campaign needs a Shopify discount owner only when the published configuration requires one.

Step 8: Preview and publish

The preview shows the desktop and mobile rendering, the API response, the current candidate pool, the expected fallback, and any conflicts with other rules. Publish when the preview matches what you expect.

How rules resolve

On the product page, the single highest-priority matching rule wins. On the cart page and cart drawer, all matching rules contribute candidates, duplicates are removed, and each recommended product keeps exactly one owning rule—so attribution never double-counts.