One offer in the cart drawer: cross selling where the shopper can still say no

Attach logic beats best seller lists

The offer should come from co purchase data: given what is in this cart, what did similar carts also contain. That data exists in any store with a season of orders, and it produces suggestions with a reason the shopper can feel: the refill for the device, the care product for the fabric, the second half of a set. A generic best seller in the drawer asks the shopper to start a new decision. An attached product asks them to complete the current one, which is a much smaller ask at this point in the journey.

Set a price ceiling as a share of cart value. An add on priced near the cart total reopens the whole purchase decision and endangers the main conversion for the sake of the extra. Operators use different caps, and the honest way to find yours is a test, not a benchmark copied from a blog. Start conservative. The drawer that protects the main order and adds a modest attach beats the drawer that occasionally doubles a basket and regularly kills one.

When to show nothing at all

Suppression rules are half the programme. Show no offer when the cart already holds the suggested product or its variant, when the shopper has dismissed the drawer offer once this session, or when the basket pattern says gift and speed: gift options selected, express checkout entered, a session measured in seconds. Every suppressed impression protects the offer's credibility for the sessions where it fits.

Stock honesty applies here too. Never offer what cannot ship with the main order on the same promise. An add on that splits the delivery or extends the date, announced after it is added, costs more trust than the margin it brought in. The drawer's small print is read at exactly the moment the shopper is deciding whether this store respects them.

Measure the drawer against the whole order

Attach rate alone will lie to you, because an offer can attach and still lose money by dragging checkout completion down or by discounting a product the shopper would have bought at full price next week. Measure the drawer on completed order value per session that saw it, against a holdout that did not, and include the return rate on attached items, which runs high when the suggestion was weak and the shopper felt nudged rather than helped.

Run the holdout longer than feels necessary. Drawer effects are small per order and novelty inflates the first fortnight in both directions. Stores that judge at day ten kill programmes that would have paid at day sixty, and keep programmes riding a launch bump that was already fading. Small levers need patient measurement, and this is the smallest lever with the largest audience in the store.

Everything in the drawer is read as the store's character at the moment of maximum attention: the offer, the shipping progress bar, the returns line, the absence or presence of pressure. A single relevant suggestion says this store knows its own products. Five spinning offers say this store would sell the shopper anything, including things that contradict the purchase they just made.