Customer Relationship Management Blogs by SAP
Stay up-to-date on the latest developments and product news about intelligent customer experience and CRM technologies through blog posts from SAP experts.
Showing results for 
Search instead for 
Did you mean: 
0 Kudos

You find questions and answers to the following topics.

1. Product Recommendations

2. Offer Recommendations

1. Product Recommendations

What type of recommendations and algorithms do you support with SAP Hybris Marketing Recommendation?

Answer:A set of algorithms come out of the box that can select products, re-rank recommendations and filter recommendation results. These algorithms belong to the following algorithm types:

·       Query (e.g. Top-Sellers, Top-N Interactions)

·       Association (e.g. often bought together, often viewed together)

·       Collaborative Filtering (e.g. User-Based Collaborative Filtering (Association, Cosine Similarity).

·       Influence recommendation results in a rule-based way (e.g. to push products of a certain brand)

·       Prebuilt capabilities like cyclicity analysis (e.g. items already recently bought within average purchase cycle of this product)

I don’t do this today nor do I have any data scientists in my team. How can I get started?

Answer:SAP Hybris Marketing Recommendation provides out of the box product recommendation templates (recommendation models) and recommendation scenarios.  By leveraging the templates, marketing experts can deliver recommendations with little effort, and without advanced statistical skills or needing to know in detail, which algorithms to use in the various recommendation scenarios they want to apply in the different channels.

Can recommendations be provided based on SAP Hybris Marketing interactions?

Answer:Yes, this is possible. Recommendations out of SAP Hybris Marketing Recommendation can be derived based on any interaction captured in SAP Hybris Marketing. Examples of interactions include: which product(s) a user has clicked on, liked, put into the shopping cart or which product(s) a user has removed from the shopping cart.

Can new/custom algorithms be added to the list of pre-delivered algorithms?

Answer:Yes, customers can extend the algorithm list by adding their own algorithms to SAP Hybris Marketing Recommendation.

Can multiple algorithms be combined within a recommendation model?

Answer:Yes, multiple algorithms can be combined at each step of the process (e.g. to selectproducts, re-rank recommendations and filter recommendation results).

Can recommendations be delivered on different Commerce Platforms?

Answer:SAP Hybris Marketing is well integrated in various SAP solutions, like SAP Hybris Commerce, SAP Customer Activity Repository and SAP Cloud for Customers, and further integration is in development.

However, SAP Hybris Marketing is a standalone solution so it can be integrated with non-SAP systems as well, on project basis. Having said this, Product recommendations can be provided on any Commerce Platform. As a consumer is browsing the commerce site, there are APIs to connect to SAP Hybris Marketing and pass the product recommendations to the Commerce Platform. Then it is up to the Commerce Platform to use this information and display it properly.

Is it possible to track the success of the recommendations being provided?

Answer:Yes, in SAP Hybris Marketing Recommendation it is possible to track the success of recommendation models. This is achieved based on the success metrics that are available for each recommendation model. This data includes the information on how many times a recommendation model was shown ("impressions") and how many times a customer clicked on a recommended product ("click through").

Other than the Web shop, what other channels can be used to deliver recommendations?

Answer:Recommendations can be provided in various channels by consuming the pre-delivered APIs. SAP Hybris Marketing Recommendation is currently integrated out-of-the-box with Hybris Commerce and SAP CRM Interaction Center. The Consumer Factsheet in SAP Hybris Marketing contains the recommendations provided to each consumer and can be used as a reference for marketing experts.

What are typical recommendation scenarios?

Use Case 1: Recommendations based on a specific product a customer is currently looking at and / or intends to purchase.

·       In a Web shop, a customer is looking at the product details page of a specific product “A” (e.g. a specific digital camera model). In SAP Hybris Marketing Recommendation, this is also called a “leading product”. With the help of SAP Hybris Marketing Recommendation, additional recommendations can now be shown to the customer, using the knowledge that he is looking at product “A”.

·       Another example could be in assisted channels, like a call center or in-store consulting, SAP Hybris Marketing Recommendation can be used to provide additional recommendations to call center agents or to the In-Store Salespeople enabling them to leverage the knowledge around the respective product they are currently discussing with the customer.

These use cases use association algorithms that can provide recommendation based on the following:

·         Purchases – “customers who bought this product also bought …”

·         Likes – “customers who liked this product also liked …”

·         Views – “customers who viewed this product also viewed …”

Use Case 2: Recommend additional product categories, leveraging the knowledge of the products the customer has already put into his shopping cart


In a Web shop, a customer places multiple items into their shopping cart. With SAP Hybris Marketing Recommendation, recommendations can be displayed on the shopping cart page to encourage the customer to choose additional products from other product categories that complement the items that are contained in the cart.

For this use case, association analysis enables us to map the products contained in the cart to a category level and recommend complimentary products.

Use Case 3: Make TOP-N recommendations to customers


In the Web shop, this scenario could show the most viewed products or the most sold products on the landing page, the checkout page, a category page, an empty search result page, a 404 page.

For this use case, “TOP-N” algorithms can be used to determine the top items (purchases, likes, views, etc ) of a category.

Use Case 4: Show recommendations to a customer based on the behavior of other customers with similar buying behavior (“Best friends”)


In the Web shop, a customer has placed multiple items into his cart. With SAP Hybris Marketing Recommendation, recommendations can be displayed on the checkout page to encourage the customer to choose additional products that other people with similar buying behavior have purchased. For this use case, algorithms of the type “Collaborative Filtering” can be used, to determine products bought by people behaving just like you, but not yet bought by you (collaborative filtering).

2. Offer Recommendations

What is the goal of providing offer recommendations?

Answer:With SAP Hybris Marketing Recommendation, offer recommendations can be provided that would enable marketers to choose, in real-time, for every customer, in every sales channel (omni-channel), in the right context, the best of all available promotional offers.

What is a promotional offer? A promotional offer is conditional promise that a seller makes to a consumer in combination with an incentive. This could be for example “If you buy product A and product B today, you get 2$ off on product B”, or “If your shopping cart value is above 100$, you do not need to pay shipping costs”, or “If you buy this cell phone together with a mobile contract, you will receive a 10$ coupon that you can use for your next purchase”.

A promotional offer, however, can also be without incentive(“incentiveless”). This is the case when marketers want to bring certain product(s) to the awareness of the customer in order to push sales (notion of offering).

What is not the goal of providing offer recommendations?

Answer:When creating offer recommendation models it does not include the creation and the maintenance of an offer. This has to be done in another solution. For example, depending on the system(s) a customer is using, this could be done in the SAP Promotion Management for Retail (SAP PMR), SAP Hybris Marketing, or a 3rd party system that the customer has in place.

In addition to that, SAP Hybris Marketing Recommendation is also not responsible for displaying the offer in the respective sales channel. This is done by the respective frontend application, for example SAP Hybris Commerce Web Content Management System for digital commerce channels.

What is the difference between a product recommendation and an offer recommendation?

Answer:Withproduct recommendations marketers are enabled to make context aware, intelligent product recommendations in real-time, across all sales channels, without providing incentives to a customer. For example, “customers who bought product A, also bought product B”, “customers who have viewed product C, also viewed product D, “customers who liked product A also like product B”.

Withoffer recommendations marketers are enabled to choose, in real-time, for every customer, in every sales channel (omni-channel), in the right context, the best of all available promotional offers.  For example, “If you buy this mobile phone together with a mobile contract, you will get a 10% discount on the mobile phone” or “If you order today online, no shipping fees are charged to you”.

How do product recommendations and offer recommendations complement each other?

Answer:For product recommendations they look at the past behavior of customers, and their peer groups and predict what will likely appeal to customers – without providing incentives. For offer recommendations the most appropriate offer out of an existing offer base (which typically has incentives) is recommended.

How a customer uses product recommendations and offer recommendations ultimately depends on a marketing strategy and what they want to achieve.

Here are some examples of how product and offer recommendations could be used to complement each other:

In a web channel, customers can decide to include either product recommendations or offer recommendations at various locations within the Web shop. Depending on their strategy, a customer may decide to provide offer recommendations on the landing page rather than product recommendations. However, on the product details page, the customer may decide to put the focus on product recommendations rather than offer recommendations.

On an empty search page, a customer may decide to mix and display both.

Where/how can we define offers? Do we need a separate solution for this?

Answer:In an ideal SAP scenario, offers are created in the SAP Promotion Management Retail (SAP PMR) or in the new offer UI called “Management Promotional Offers” and then replicated into SAP Hybris Marketing's central offer repository. In SAP Hybris Marketing a shadow offer object will then be created, in order to carry the digital content for the respective offer (e.g. image, promotional text …)

In addition, it is also possible to create an offer in a 3rd party system. In this case, the offer needs to be replicated into SAP Hybris Marketing so that a shadow object is created in the SAP Hybris Marketing Offer repository that will then contain the digital content.

The creation of the shadow object in the SAP Hybris Marketing Offer repository is possible, as SAP Hybris Marketing Offer provides an API (RESTful service) that allows the creation of such an object.

If there is no price-relevant information attached to an offer, it is also possible to simply create the offer in SAP Hybris Marketing. It will then only contain the digital content, but no price related details.

How are offers recommended?

Answer:In SAP Hybris Marketing Recommendation, the marketer defines, maintains and manages offer recommendation scenarios and offer recommendation models. An offer recommendation model works based on rules. The marketer creates rules for selecting the most suitable promotional offers from the offer database, based on the respective contextual situation. These rules are composed in Offer Recommendation with easy to understand rule fragments, using a graphical rule editor. A rule fragment is a pre-defined expression that is used to select, re-rank, or to filter offers out of a data pool.

SAP Hybris Marketing Recommendation comes out of the box with multiple rule fragments. In addition to that there is also an extension concept in place, so customers can add new rule fragments. In addition to that preconditions like e.g. target groups can be defined, and like this target group specific recommendation models can be build.

Examples for a rule fragment are “choose offers including the currently viewed product” or “choose offers including the specific offer ID 1234” or “choose offers for target group young urban families”.

Before activating a recommendation model, the marketer can also preview the recommendations returned by an offer recommendation model.

For which sales channels can offer recommendations be done?

Answer:SAP Hybris Marketing Recommendation is OMNICHANNEL

SAP Hybris Marketing Recommendation can be called from ANY sales channel and like this can provide at all customer touch points a consistent and context related shopping experience for the end-customer in alignment with a customer’s marketing strategy.

SAP Hybris Marketing Recommendation can be consumed in sales channels like in-store service environments, point-of-sale, consumer mobile apps, call center, E-Mail campaigns, and on the web.

How can proximity based marketing be realized with offer recommendations?

Answer:Proximity based marketing can be achieved by using triggers from a mobile device or from another IoT related technology like, for example, indoor positioning, iBeacon, etc.

To initiate the recommendation of an offer, the offer information needs to have location information to be able to filter the relevant offers by proximity that come from the trigger point of the customer. For example, the location information could be a “store ID” to identify the relevant stores, for which then the relevant offers can be derived.