Tab Commerce

Strategies and optimization

How Fixed order, Rotate and Optimize decide which of your approved add-ons show, what the optimization goals mean, and how to keep manual control.

Every Upsell rule runs one of three strategies. Bundles do not use this allocator. The strategy decides how the candidates you approved share the cards a placement shows; it never adds a product you didn’t pick.

Fixed order

Candidates show in the order you set, every time, with pinned candidates first. Use Fixed order when you already know the priority, when traffic is too low for optimization to learn anything, when a campaign has just launched, or when specific products must always be visible. One candidate is enough.

Rotate

Pinned candidates keep their place. The remaining slots go to the candidates with the fewest views so far, with ties broken at random, so views spread as evenly as possible across the pool. Use Rotate to keep a lineup fresh, or to collect even initial data before switching a rule to Optimize. Rotate needs at least two candidates.

Optimize

Optimize, on the Optimize and Scale plans, shares the slots according to measured performance against the goal you choose. It needs at least two candidates.

  • Exploration is permanent. 10% of selections are always spread evenly, so a candidate that starts slowly can still earn its way in.
  • New candidates get a fair read. Until a candidate has 100 valid views, it gets an even share.
  • No candidate takes over. Each candidate’s share stays between 5% and 50%.
  • Recent results count most. Optimize looks at the last 7 days, with older results fading by half every 3 days, and recalculates about every 10 minutes.

Optimization goals

  • Add to cart rate — best for low-consideration add-ons, especially in the cart.
  • Purchase rate — best when adding is easy but purchases are what matter.
  • Revenue per impression — best when candidate prices differ widely and you care about value per slot.
  • Attributed gross profit per impression (Scale) — uses product costs from Shopify. A candidate’s weight changes only after it has at least 5 attributed orders with known costs, and those costs cover at least 90% of its sold quantity and revenue. Until then it shows as low cost coverage or insufficient profit data.

On Scale you can also set an Observed gross margin floor. Candidates whose observed margin falls below it drop back to the exploration share instead of competing for more views.

Plan access and placement scope

Fixed order and Rotate are on every plan. Optimize needs the Optimize or Scale plan, and the gross-profit goal and margin floor need Scale. See Plan access and analytics.

A rule can target several placements. Its candidate pool is shared across those placements, while events and reports keep the placement where the action happened. Create separate rules when you want different products, or independent optimization, in each placement.

How rules combine

  • Product page: only the highest-priority matching rule shows.
  • Cart page and cart drawer: every matching rule contributes, taking turns, with duplicates removed and up to 20 cards in total.

A product already in the cart is never recommended, and a product never triggers a recommendation of itself.

Manual control inside Optimize

Optimization never removes your controls. You can remove a candidate from the pool, pause or resume optimization, reset what a rule has learned, or switch the rule back to Fixed order or Rotate at any time.