The eight strategies
Related — products similar to what's in the cart: same category, same style. The general-purpose cross-sell.
Complementary — products that go with cart items rather than resembling them: accessories, add-ons, the consumable that fits the hardware.
Subscription upgrade — finds cart items that have a subscription plan available and offers the subscription version of the one-off purchase.
Size upgrade — offers the next-larger variant of something already in the cart. The price difference is calculated on the shopper's own market prices, so “upgrade for $12 more” is correct in every currency.
Same product — another unit of a cart item at a discount. The simplest volume play, and the lowest-friction post-purchase offer there is.
Clearance — surfaces low-inventory variants, with discounting that escalates automatically (below).
Rewards threshold — picks a product that closes the gap to the next tier of a rewards program. Rather than telling a shopper they're $14 from free shipping, it offers something that costs about $14.
Recently-viewed recovery — offers products the shopper looked at during this session but never added to the cart.
Clearance: markdowns that escalate on their own
Clearance is the most automated of the eight. You define what counts as low inventory, either as a unit count (on-hand at or below a number) or as days of cover — on-hand divided by recent sales velocity, which adapts to how fast each variant actually moves. Variants with no recent sales fall back to the unit check.
You then define a discount ladder. A variant starts at the first step, and advances to the next one when it accumulates a set number of impressions at its current discount without converting. Stock that shifts at 10% off never gets marked down further; stock that doesn't move walks itself down the ladder until it does.
Where recommendations start from
ShopBrain needs a seed product to reason from, and you control where it looks — as an ordered fallback chain. The options are the cart, the product being viewed, the shopper's recently viewed products this session, and the last order of a logged-in customer. The first source that yields a product wins; sources with nothing available are skipped.
That ordering is a real lever. “Viewed product, then cart” behaves like a classic product-page cross-sell. “Last order, then cart” behaves like replenishment for returning customers. On a surface with no cart and no product in view, the chain simply yields no seed and configured fallback products still apply.
When several cart items are eligible, you choose whether to upsell from the highest-value item or the first eligible one.
Staying inside your catalogue
Every strategy is constrained by filters you set. Include by collection, product tag or product type — all of which must match, so you can narrow precisely. Exclude by the same three, where matching any one rule removes the product.
This is what makes automated recommendation safe to run: a gift-card collection, discontinued lines, or anything you never want discounted can be excluded once and stays excluded no matter which strategy is running.
You also set how many products a recommendation returns — between one and ten, two by default — and recommended products can carry their own discount, since they have no individual product entry to attach one to.
It changes the products, not the rules
A ShopBrain offer is still an ordinary offer. Its targeting, behavior mode, placements, discount and reporting work exactly as they would with a hand-picked product list. The only thing that changes is which product fills the slot for a given shopper. You can also write a message template with dynamic tokens so the copy adapts to whatever was selected.
Configuration is per offer, not account-wide, so two ShopBrain offers can run entirely different strategies at the same time.
