Next-basket prediction

A plain recommendation engine answers one question: which products are similar to this one. Next-basket prediction answers a different one: which products is this person likely to put in her next order. For a grocery or a pet store the two answers differ a lot. The shopper who bought cat food three weeks ago needs cat food again, not a cat toy that looks similar.

Why it matters for a store

Repeat purchase is where margin lives in grocery, pharmacy, pet supplies, beauty and supplements. A prediction of the next basket lets the store remind the right people at the right moment and fill the cart page with items they actually meant to buy. It is one of the 16 named algorithms the agent runs, next to recommendations by similarity and by session.

An example

An online grocery shows each shopper, before checkout, the items she usually buys and has not added yet. In that store 22% of all orders included an item from that cart block, and the block earned 68% of all recommendation revenue, as our use case on the cart page shows.

How the agent uses it

The sales agent uses next-basket prediction on the cart page and in recommendations; retention uses it to time replenishment messages. How it decides explains how the goal you choose changes what the model is asked to optimize.

Related terms

See also RFM segmentation, Predicted customer lifetime value, Abandoned cart.

Where it shows

Product blocks on the home, category, product, cart and thank-you pages. For each visitor and page the agent picks the approach, such as next basket or size and color match, tests choices and placements against each other, and moves traffic to what sells. Your own rules and product attributes sit on top: nobody has to pick a winner by hand.