CX Works

CX Works brings the most relevant leading practices to you.
It is a single portal of curated, field-tested and SAP-verified expertise for SAP Customer Experience solutions

Personalized Recommendations - How to Get Started

23 min read

Overview

Personalized Recommendations - How to Get Started

Consumers are willing to share personal data for a clear personal benefit. Recent surveys show that around 80% of consumers are willing to share basic personal information for personalization. The same survey shows that personalization is a critical factor in earning customer loyalty.

Personalization exists in different levels, from simple "Dear [First Name]" email personalization via condition-based content to personalized recommendations. This article explains how to get started with recommendations in SAP Marketing Cloud. After reading this article, you can best continue with the article Recommendations - Set Up and Operate to learn how you should set up and operate recommendations within SAP Marketing Cloud. 

When talking about recommendations, we are typically referring to product recommendations. The result of a product recommendation scenario is a ranked list of products. However, the concept of recommendations can easily be applied to further scenarios, e.g. to recommend offers, brands, or e-learning content. Most best practices and guidelines included in this article focus on product recommendations, but apply to other scenarios as well.  

Table of Contents

How to Get to a Customer-Specific Roadmap

Before starting with recommendations, working on a roadmap tailored to requirements and goals is critical to ensure success. This article explains the most important steps which are necessary to get to a customer-specific roadmap for recommendations. 

The recommendation scenario roadmap will answer the two main questions what do you want to do? and - more importantly - why do you want to do it? 

image one: scenario roadmap - what and why? 

Please notice that the exact plan around implementation details, such as the number of recommendation results, algorithm parameters, or weights are not discussed in this article.

You can find these in the next article Personalized Recommendations - Set Up and Operate

Discuss the Goals Across your Buyer's Journey

We found that the buyer's journey stages help to plan your recommendation roadmap. If you do not have your buyer's journey stages sorted out yet, you might want to use the four stages as described in our article Business Scenarios - Learn How to Articulate Tangible Campaigns. The four stages Awareness & Interest, Consideration, Conversion, and Retention & Advocacy typically serve the purpose since they group recommendation scenarios with similar goals together. 

image two: buyer's journey stages

Before looking into recommendation scenarios, you should discuss your buyer's journey and the corresponding metrics to measure business success. This is crucial for evaluating the effectiveness of your scenarios and fine-tuning the models with respect to these metrics later on. 

Examples of business success metrics that are tracked in SAP Marketing Cloud are:

  • Click-Through Rate
    • The proportion of recommended products that are clicked by consumers.
  • Conversion Rate
    • The proportion of recommended products that are purchased by consumers.
  • (Impressions)
    • The number of times a scenario has provided product recommendations to a consuming application.
      Please notice that this metric typically does not represent business success with regards to recommendations since the metric only depends on where and how recommendations are presented in the consuming application.

Based on your scenarios, there might be other, externally tracked metrics which might be of interest.

The following table shows how typical metrics by Buyer's Journey Stage can look: 

Buyer's Journey Stage Typical Metrics
Awareness & Interest

Impressions

Unique Visitors

Consideration

Product Page Views

Link Clicks

Conversion

Conversion Rate

Revenue

Retention & Advocacy

Engagement Scores

Net Promoter Score

Discuss your buyer's journey and the corresponding metrics to measure business success. 

Your custom buyer's journey and metrics will look different, but the examples above should give you an idea. 

Leverage the Recommendation Scenario Catalogue

The next step is to find recommendation scenarios that support your metrics across your buyer's journey. 

Each recommendation scenario generates recommendations differently based on the context, related data, and the underlying algorithms

Over the last years, we collected recommendation scenarios from customers doing B2B and B2C business across various industries in a Recommendation Scenario Catalogue. In the next step, we grouped similar recommendation scenarios according to their fit into the buyer's journey stages: 

  • These recommendation scenarios focus on the beginning of the buyer's journey. Here is where consumers with an early interest in your offerings have their first touchpoint with your brand and begin to search and research. As you might expect, these scenarios are contained in early-acquisition campaigns (e.g. Welcome/Onboarding campaigns) or at the Homepage of the webshop.

  • Many recommendation scenarios belong to the Consideration buyer's journey stage. In this stage, the customer might already have done some early research and evaluation of alternatives, but no decision has been made yet. This research already allows to better understand the consumer's interest to build a consumer profile. The scenarios in this stage are mostly based on the product or product category context.

  • Recommendation scenarios belonging to the Conversion buyer's journey stage are at the purchase stage of the cycle. That's why recommendations at this stage are typically placed on the cart or order confirmation pages for cross and up-selling to increase the order value. 

  • Depending on which industry you are in and what you're selling, recommendation scenarios in the Retention & Advocacy can look very different. Recommendations scenarios focusing on Retention & Advocacy might try to make customers repurchase or continuously nurture the relationship with them. 

  • Apart from the fit into the buyer's journey stages, some recommendation scenarios are rather generic and - depending on the implementation or usage - might fit into different journey stages or in multiple stages at the same time.

Apart from the previous groups which typically generate recommendations, the next group includes approaches or algorithms to refine the recommendation scenarios. 

  • Approaches or algorithms which are taking the results from the previous groups and further filter or post-process these can be clustered into this group.  

As explained earlier, the examples mainly focus on product recommendations. However, in a lot of cases, you can also build your recommendations for other entities like offers, events or content. As an example, for learning-content recommendations, you could build your recommendation algorithms on the conversion interaction enrollment to e-learning instead of on product purchase.

Find recommendation scenarios that support those metrics you identified in the last section.

You can leverage the Recommendation Scenario Catalogue to find recommendation scenarios or to get inspiration for custom scenarios.

Create a Scenario Roadmap

After identifying recommendation scenarios, you are ready to create a roadmap for recommendations. 

Identical to other business scenarios and features, the same underlying principles should be applied when creating a roadmap for recommendations. Based on our experience in various regions and industries, the "crawl, walk, run" approach leads to higher value realization, project success, and adoption. 

We recommend following the progressive "crawl, walk, run" approach for implementing and rolling out recommendations in SAP Marketing Cloud.

This approach should be followed for all recommendation-related aspects, such as the selection of recommendation scenarios, the underlying algorithms, or consuming channels. 

Find more information in the following article: Master the Challenges of Digital Marketing.

We recommend that you focus on one or two journey stages per phase and that you select one or two scenarios for those channels which are most relevant for you. Your selection criteria should identify which recommendation scenarios best fit to your measurable metrics and high-level objectives.

An example of a recommendation scenario roadmap is displayed below: 

Phase Journey Stages Scenarios (E-Mail) Scenarios (Webshop) Metrics High-level Objective
1

Awareness & Interest

Consideration

  • Top selling products in Onboarding/Welcome campaign
  • Last viewed products in newsletters or retargeting campaigns
  • Top selling products on Homepage

Impressions

Product page views

Apply simple standard recommendation models to gain the first experience and build knowledge.
2 Conversion
  • Cross-sell in after-sales campaign (survey, rating, thank you)
  • Cross-sell of complementary products (within/across session) on
    • Product detail page
    • Cart page
    • Order confirmation page

Conversion rate

Revenue

Apply standard cross-sell scenarios and measure marketing effectiveness.
3

Consideration

Conversion

  • Last purchased product in replenishment campaign
  • Complementary product categories in newsletters/after-sales campaigns
  • Include post-processing logic (out of stock, already purchased)

Product page views

Conversion rate

Revenue

Collect experience with a simple custom algorithm to enable use cases going beyond simple product cross-sell
4 Conversion
  • Complementary products ranked by product profitability
  • Dynamic offer recommendation

Revenue

Offer conversion

Optimize marketing contribution with custom algorithms

Create a roadmap for recommendations to answer what you want to do in each phase.

Also, the recommendation roadmap can help you to answer why certain recommendation scenarios help to achieve your metrics and objectives per phase and journey stage. 

Plan the Next Steps

The recommendation scenario roadmap serves as the basis for the next steps. Documenting the selected recommendation scenarios in more detail is important before you start with the actual implementation. The scenario details will answer the question how do you want to implement the recommendation scenarios.

image three: scenario details - how? 

All dimensions and the level of detail required for the recommendation scenarios are described in the next article Personalized Recommendations - Set Up and Operate.

Create the recommendation scenario details to document how you want to implement the recommendation scenarios. Find more information in the article Personalized Recommendations - Set Up and Operate.

Conclusion

Based on what you have learned within this article, maybe you want to try out recommendations in SAP Marketing Cloud and start building your own recommendation roadmap. 

This article explained how to get started with recommendations in SAP Marketing Cloud. The main steps have been outlined as follows: 

  • Discuss the metrics across your buyer's journey
  • Leverage the Recommendation Scenario catalogue
  • Create a scenario roadmap
  • Plan the next steps

If you're interested in learning more about the next steps, you should continue with the article Personalized recommendations - Set Up and Operate. This article contains various practical examples and recommended practices to avoid common challenges during the set up and operations phase. 

Appendix: Recommendation Scenario Catalogue

The Recommendation Scenario catalogue is available for download here and included below.

Awareness & Interest

Scenario

Context

Section

E-Mail Campaigns

Industry-specific

Effort

Algorithm

Additional Data (**)

Top sellers

-

  • Homepage
  • Welcome/Onboarding

All

Very low

Standard, Top Sellers (Interactions)

  • Sales orders / transactions

Top viewed / trending products

-

  • Homepage
  • Welcome/Onboarding

All

Very low

Standard, Top Viewed

  • Product views

Discounted / promotion products

-

  • Homepage
  • Welcome/Onboarding

All

Medium

Custom algorithm

  • Offers / promotions

High-rated products

-

  • Homepage
  • Welcome/Onboarding

All

Medium

Custom algorithm

  • Product ratings

New arrivals / Just arrived

-

  • Homepage
  • Product launch

All

Medium

Custom algorithm

  • Products (w/ entry date)

Targeted top sellers
(e.g. in country / market, by gender …)

  • Contact
  • Homepage
  • Targeted communication

All

Low

Standard, Top Sellers (Interactions) (*)

  • Products
  • Sales orders / transactions

Bought together with installed base products

  • Contact
  • Homepage
  • Targeted communication

Utilities, Industrial Machinery, High Tech …

High

Custom algorithm

  • Products
  • Installed base data
  • Sales orders / transactions

Consideration

Scenario

Context

Section

E-Mail Campaigns

Industry-specific

Effort

Algorithm

Additional Data (**)

Top sellers in product category

  • Product category
  • Category pages
  • Product list pages
  • Category or interest based communication

All

Low

Standard, Top Sellers (Interactions) (*)

  • Sales orders / transactions
  • Product category catalog

Top viewed in product category

  • Product category
  • Category pages
  • Product list pages
  • Category or interest based communication

All

Low

Standard, Top Viewed (*)

  • Product views
  • Product category catalog

Top rated in product category

  • Product category
  • Category pages
  • Product list pages
  • Category or interest based communication

All

Medium

Custom algorithm

  • Product ratings
  • Product category catalog

Similar products

  • Product
  • Searched Product
  • Product detail pages
  • 404 / no results page
  • Retargeting

All

High

Custom algorithm

  • (only product data)

Other products bought after viewing

  • Product
  • Product detail pages
  • Retargeting

All

Medium

Standard, Other Items Bought After Viewing

  • Sales orders / transactions
  • Product views

What similar customers bought …

  • Contact
  • Category pages
  • Product list pages
  • Newsletter

All

High

Standard, User Interaction Based Collaborative Filtering

  • Sales orders / transactions
  • Product views

Recommend based on Interest

  • Contact
  • Homepage
  • Interest-based communication

All

Low

Standard, Recommend Based on Items of Interest

  • Interactions with item of interest data

Last viewed / abandoned products

  • Contact‘s last viewed products
  • Homepage
  • 404 / no results page
  • Retargeting
  • Back in stock

All

Low

Standard, Recently Viewed Items

  • Product views

Bought together with last viewed products

  • Contact‘s last viewed products
  • Homepage
  • Retargeting
  • Back in stock

All

Medium

Standard, Often Bought Together (Interactions) (*)

  • Sales orders / transactions
  • Product views

Recommend wish list products

  • Contact‘s wish list products
  • Homepage
  • 404 / no results page
  • Newsletter
  • Wish list reminder campaigns

Consumer products

Medium

Custom algorithm

  • Wish list data

Bought together with wish list products

  • Contact‘s wish list products
  • Homepage
  • Newsletter
  • Wish list reminder campaigns

Consumer products

Medium

Standard, Often Bought Together (Interactions) (*)

  • Wish list data
  • Sales orders / transactions

Conversion

Scenario

Context

Section

E-Mail Campaigns

Industry-specific

Effort

Algorithm

Additional Data (**)

Often bought together

  • Product
  • Products in cart
  • Products purchased
  • Product detail pages
  • Cart page
  • Order confirmation page
  • After-sales (survey, rating, thank you)
  • Retargeting

All

Medium

Standard, Often Bought Together (Interactions) (*)

  • Sales orders / transactions

Complete the collection / look / kit

  • Product
  • Products in cart
  • Products purchased
  • Product detail pages
  • Cart page
  • Order confirmation page
  • After-sales (survey, rating, thank you)
  • Retargeting

All

Medium

Custom algorithm

  • Collection / look / kit

Retention & Advocacy

Scenario

Context

Section

E-Mail Campaigns

Industry-specific

Effort

Algorithm

Additional Data (**)

Buy again
(e.g. recurring purchases)

  • Contact‘s purchased products
  • Homepage
  • Replenishment

Retail / Consumer

Medium

Standard, Last Purchased Items

  • Sales orders / transactions
  • Replenishment period

Generic Approaches

Scenario

Context

Section

E-Mail Campaigns

Industry-specific

Effort

Algorithm

Additional Data (**)

Position a product

-

-

-

All

Low

Standard, Position a Product

-

Generic external recommendation

-

-

-

All

Low

Standard, External Product Recommendation List

  • External recommendation data

External recommendations based on contact

  • Contact

-

-

All

Low

Standard, External Product Recommendation List

  • External recommendation data

External recommendations based on product

  • Product

-

-

All

Low

Standard, External Product Recommendations Based on Leading Items

  • External recommendation data

External recommendations based on contact and product

  • Contact
  • Product

-

-

All

Low

Standard, External Product Recommendations Based on Leading Items

  • External recommendation data

Rule-based (if-this-then-that)
[e.g. for this product category / brand show this product category / brand]

  • Depending on implementation

-

-

All

Medium-High

Custom algorithm

  • Depending on implementation

Filter & Post-processing

Approach / Algorithm

Context

Section

E-Mail Campaigns

Industry-specific

Effort

Algorithm

Additional Data (**)

Remove products already purchased

  • Contact‘s purchased products

-

-

All

Low

Standard, Remove Items Already Purchased

  • Sales orders / transactions

Remove products from product categories already purchased

  • Contact‘s purchased products

-

-

All

Low

Standard, Remove Item Categories Already Purchased

  • Sales orders / transactions
  • Product category catalog

Remove products already in cart

  • Products in cart

-

-

All

Low

Standard, Remove Items Already in the Cart or custom algorithm (*)

  • Cart products

Remove products from product categories already in cart

  • Products in cart

-

-

All

Low

Standard, Remove Item Categories Already in the Cart or custom algorithm (*)

  • Cart products
  • Product category catalog

Remove out of stock products

-

-

-

All

Medium

Custom algorithm (*)

  • Product w/ stock information

Remove leading products

  • Product

-

-

All

Low

Standard, Remove Leading Items

  • Leading products

Re-rank products

  • Manipulate the recommendation results and change the order, e.g. by product margin

-

-

All

High

Custom Algorithm, CX Works Example

  • Product Margin
Overlay