Intelligent Selling Services in SAP Commerce Cloud: AI-Powered Personalisation

Your customers don’t think about “personalisation.” They don’t care about your recommendation engine. They just leave when the products on screen feel random. That’s the whole problem in one sentence. Intelligent Selling Services (ISS) is Commerce Cloud’s answer to it. It’s the AI layer that tunes product recommendations, search results, and merchandising to each shopper’s behaviour.

We’ll walk through what ISS does, how the four core capabilities work, and what kind of lift you can realistically expect.

TL;DR: SAP has over 34,000 AI customers across its portfolio, with plans to reach 100,000 by 2027 (SAP News, 2024). Intelligent Selling Services (ISS) is Commerce Cloud’s AI personalisation engine — it handles product recommendations, adaptive search, dynamic merchandising, and context-driven personalisation. ISS analyses clickstream and purchase data to show each customer the products most likely to convert.

What Is Intelligent Selling Services?

SAP has over 34,000 AI customers with plans to reach 100,000 by 2027 (SAP News, 2024). ISS is one of the ways that shows up in Commerce Cloud.

The thing to understand: ISS isn’t a recommendation widget you bolt on. It’s a platform-level cloud service that analyses customer behaviour (clicks, searches, cart additions, purchases) and personalises four aspects of the shopping experience at once:

1. Product recommendations — “Customers who bought X also bought Y”
2. Adaptive search — search results ranked by individual relevance
3. Dynamic merchandising — category pages that reorder products per customer
4. Context-driven personalisation — content and offers tailored to browsing context

ISS runs separately from the Commerce Cloud application server. It ingests behavioural events from the storefront, processes them through ML models, and returns personalised results via API. That separation matters: ISS crunches data without slowing down your storefront. Neat architecture.

How Do Product Recommendations Work?

Global retail e-commerce reached $6.334 trillion in 2024 (eMarketer, 2024). Product recommendations directly influence what share of that spending goes to your store versus someone else’s.

ISS gives you several recommendation strategies. Each earns its keep in a different scenario:

Collaborative filtering. The classic “customers who bought this also bought…” approach. Draws on purchase patterns across your whole customer base. More transaction volume you have, the crisper the recommendations get. We had one client go from generic “top sellers” to collaborative filtering and see a 3x lift in recommendation click-through within six weeks.

Content-based filtering. Recommends products with similar attributes (category, brand, price range) to what the customer viewed or purchased. Works even with limited purchase history. A first-time visitor who browses three products already gets decent suggestions.

Context-aware recommendations. What you recommend changes depending on where the customer is in their journey. A recommendation on the cart page (cross-sell) is a very different beast from one on the homepage (discovery). ISS adapts accordingly.

Trending products. Surfaces products with rising demand. Useful for seasonal items, new launches, or products that suddenly catch fire.

The engine runs continuously, updating models as new data arrives. You don’t retrain anything manually. ISS handles that on its own.

How Does Adaptive Search Differ from Standard Search?

Gartner named SAP a Leader in Digital Commerce for 11 consecutive years (SAP News, 2025). ISS’s adaptive search is one of the reasons.

Standard Solr search gives every customer the same results for the same query. Fine for a catalogue. Not fine for a store trying to convert. Adaptive search adds a personalisation layer:

Personalised ranking. Two customers searching “running shoes” see different products first. One gets trail shoes because they’ve browsed outdoor gear. The other gets road shoes because they’ve viewed marathon content. Same query, different intent, different results.

Behavioural boost. Products the customer previously viewed, carted, or purchased in the same category get a push in search results. Not to the point of filter bubbles, but enough to surface the most relevant options first.

Merchandising rules. Your merchandisers set rules that interact with adaptive search: boost new arrivals, promote high-margin products, suppress out-of-stock items. These combine with the AI-driven personalisation. Human judgement and machine learning working together.

Category page optimisation. Something people miss: adaptive search doesn’t just affect the search box. It also reorders category pages. When a customer browses “Men’s Shoes,” products are ranked by predicted relevance to that specific customer. I reckon this is where more revenue hides than most commerce teams realise.

What Does Dynamic Merchandising Do?

E-commerce accounts for 34% of B2B revenue globally (McKinsey, 2024). Dynamic merchandising makes sure revenue-driving products are visible to the right customers.

This is where ISS earns most of its keep in practice:

Automated product sorting. Category pages and search results get sorted by predicted conversion probability for each customer. Products that customer is most likely to buy float to the top. No manual curation needed (though you can override).

Seasonal and trend-based adjustment. ISS picks up on rising demand patterns. Winter jackets get boosted as temperatures drop. Trending products rise in category rankings. Nobody has to flip a switch.

Inventory-aware placement. Low-stock products can be de-prioritised to avoid overselling. Products sitting heavy in the warehouse get a boost to accelerate sell-through. That’s a proper operational win, not just a UX one.

A/B testing support. Run experiments: compare AI-driven sorting against manual curation to validate that ISS actually improves conversion for your specific catalogue and customer base. Trust but verify.

What Results Can You Expect?

61% of B2B buyers prefer a rep-free buying experience (Gartner, 2025). When customers self-serve, AI-driven personalisation has to do the job a good sales rep used to do.

ISS impact varies by catalogue size, traffic volume, and how mature your personalisation was before. Here’s what we typically see:

• Click-through rate on recommendations: 2-5x improvement vs. static “top sellers” lists
• Search-to-cart conversion: 10-20% improvement with adaptive search vs. standard Solr
• Average order value: 5-15% increase through contextual cross-sell recommendations
• Time to find products: measurable reduction when adaptive search surfaces relevant results faster

Fair enough, these aren’t guaranteed. They depend on data volume, catalogue depth, and customer behaviour patterns. ISS needs traffic to learn. Low-traffic stores see smaller gains because the models have less to work with.

FAQ

Is ISS included with Commerce Cloud?

Yes, ISS is part of SAP Commerce Cloud. Which features you get depends on your edition and licensing, so check with SAP or your implementation partner for specifics.

How much traffic does ISS need to be effective?

ISS needs meaningful behavioural data to train its models. Stores with fewer than 1,000 monthly sessions will see limited personalisation quality. The good news: the engine improves continuously as more data arrives.

Can I control what ISS recommends?

Absolutely. Merchandisers can set business rules that override or influence ISS: boost brands, suppress products, set minimum margin thresholds, pin products to specific slots. AI and manual curation work together. Exactly how it should be.

Does ISS work with the Composable Storefront?

Yes. ISS exposes capabilities through APIs that the Composable Storefront consumes. Recommendation carousels, personalised search results, dynamic category sorting. All available through standard components.

How does ISS compare to third-party recommendation engines?

ISS has one solid advantage: deep integration with Commerce Cloud. Same product catalogue, same customer data, same order history. Third-party engines (Algolia Recommend, Nosto, Dynamic Yield) may have more sophisticated algorithms or broader cross-platform support, but they need extra integration work and data sync. Prima vista, ISS wins on simplicity; third-party engines win on flexibility.