Generic recommendations ("customers also viewed") have lost effectiveness. Modern AI crosses individual history, margin, inventory, and journey stage to suggest what the shopper is most likely to buy now.
Session context
First visit, return within 24h, or repeat buyer require different logic. Use real-time signals — browsed category, cart items, acquisition campaign — to adjust product ranking.
Where to show recommendations
PDP, cart, post-visit email, and WhatsApp perform differently. Test placement and format; the same recommendation does not work equally across channels.
Avoid margin cannibalization
Include business rules: prioritize items with healthy stock and positive contribution. Klivvo syncs orders and events to power segments and campaigns with relevant recommendations.