Recommendations that know what happens next
The nexCommerce recommendation engine decides the next best product, offer and action for each shopper in real time, using behaviour, purchase history, loyalty tier, location, live stock and margin rules. It powers the storefront, the mobile app, campaigns and the AI shopping assistant from one model, so a customer sees one coherent set of suggestions wherever they are.
Key takeaways
- Next best product, offer and action, computed per session and refreshed on every signal.
- Delivery-aware: only recommends what can reach the customer's address and slot.
- Loyalty-aware: member prices, tier rewards and streaks feed the ranking.
- One engine serves storefront, app, push, email, SMS and the assistant.
What is next best action in ecommerce?
Next best action is the single most useful thing to put in front of a customer right now — a product, an offer, a reminder or nothing at all. It differs from a recommendation carousel because it is chosen against the whole context: what they have bought, what is in stock near them, what they can actually receive, and what your margin rules allow.
Where do recommendations appear?
One engine, every surface, so the suggestion a customer sees in an email matches what they find when they open the app.
| Surface | Placement | Example |
|---|---|---|
| Storefront | Home, category, product, cart, checkout, post-purchase | "Frequently bought together", "Complete the look", "Ships today near you" |
| Mobile app | Home feed, search, cart | Reorder suggestions, streak rewards |
| Campaigns | Push, email, SMS, banners | Segment-level offers with individual product slots |
| Assistant | Conversation | "Because you bought X, this Y fits and arrives Saturday" |
| B2B | Buyer portal | Contract-priced replenishment suggestions |
How does the recommendation engine rank?
Candidates come from collaborative signals, content similarity and merchandising rules. They are filtered by availability at the shopper's location and slot, then ranked by predicted conversion and the business weights you set — margin, stock age, campaign priority. Merchandisers can pin, boost or exclude, and every rule is auditable.
- 01Collect signals: views, searches, carts, orders, loyalty events, location.
- 02Generate candidates from behaviour, content and merchandising rules.
- 03Filter by stock, delivery feasibility and eligibility — age, region, B2B contract.
- 04Rank by predicted value with your business weights applied.
- 05Serve, measure, and feed outcomes back so the next ranking is better.
What are delivery-aware recommendations?
A recommendation for something that cannot reach the customer in their chosen slot is worse than no recommendation: it converts, then fails. The engine reads the same availability and slot data the storefront does, so what it suggests is what can actually arrive.
delivery orchestration softwareSee what it would recommend to your customers
The interesting question is not whether recommendations lift conversion. It is what the engine does with your stock, your slots and your margin rules.
Does personalisation respect GDPR consent?
Yes. Signals are collected under the consent state recorded for each visitor, and anonymous sessions receive context-only recommendations rather than behavioural ones.
Can merchandisers override the model?
Yes — pin, boost, exclude and rule-based slots, with reporting on what the override cost or gained.
Does it work at launch with little data?
Yes. Content-based and rule-based candidates cover the cold start, and behavioural signals take over as data accumulates.