Skip to Content
Technical Articles

Increase Customer Response Rate and Engagement by Providing Personalized Recommendations in Emails

Background

Customers today expect you to know what they need and what their interests are. Providing a personalized customer experience can increase the customer response rate and engagement, in turn, enabling you to optimize your marketing campaigns and drive conversions.

This blog describes how you can engage customers by acting upon recent purchases to drive conversions with personalized emails using one of the new recommendation algorithms delivered in the 1908 version of SAP Marketing Cloud.

Business Scenario

Alice Dubois and Gabriel Lécuyer have just made the following purchases:

.

              Alice – Red Coffee Maker                      Gabriel – Bag of Dark Chocolate

Note: For the purpose of this example, we are targeting only two customers. Obviously, a typical real scenario would involve many more customers and purchases.

You’d like to quickly respond to these purchases by sending Alice and Gabriel an email that contains personalized content. The emails contain the following:

  • A section that engages each customer by thanking them for doing business with your company.

Alice’s Email

 

Gabriel’s Email

 

  • A second section that leverages SAP Marketing Cloud Recommendations to cross-sell products and drive conversion. This section recommends additional products that were often bought by other customers when they purchased the same products as Alice and Gabriel.

For example, Alice’s email would contain the following in this section:

Implementation

Configuring the Recommendation Scenarios Required

The personalized content contained in each section of the email is supplied by two recommendation scenarios.

Recommendation Scenarios in Manage Recommendations App

The LAST_PURCHASE (QW_LAST_PURCHASE) Scenario

The LAST_PURCHASE (QW_LAST_PURCHASE) scenario uses a model, by the same name, that contains the new Last Purchased Items (Interactions – Optimized) recommendation algorithm.

LAST_PURCHASE Model in Manage Recommendations App

The Last Purchased Items (Interactions – Optimized) algorithm is configured with the following pre-filters:

Algorithm Pre-Filters in Manage Recommendations App

The algorithm retrieves SALES_ORDER interactions that have occured over the last 30 days and returns the products purchased by individual contacts in the last sales order within that period.

You can customize these settings to adapt them for a different business scenario. For example, you can change the Interaction Type from SALES_ORDER to SHOP_ITEM_VIEW and change Use Interaction Data (in days) from 30 to 15. The algorithm would then retrieve the products that individual contacts have browsed, rather than purchased, within the last session of a 15 day period.

The LAST PURCHASED CROSS (QW_LAST_PURCHASED_CROSS) Scenario

The LAST PURCHASED CROSS (QW_LAST_PURCHASED_CROSS) scenario uses a model called LAST_PURCHASED_CROSS that contains two algorithms, each contained in a seperate Recommend Products step. The results generated by the first step are used as the input of the second step.

LAST_PURCHASED_CROSS Model in Manage Recommendations App

The Last Purchased Items (Interactions – Optimized) algorithm returns the products purchased by individual contacts in the last sales order within that period, just as it did in the first scenario. The Often Bought Together (Interactions) algorithm looks at SALES_ORDER interactions that have occured over the last 60 days and returns the products often bought together with the products the algorithm was provided by the first step.

Note: Currently, the Remove Item Categories Already Purchased (Interactions) and Remove Items Already Purchased (Interactions) post-processing algorithms are not compatible with the email channel in 1908. The product management team is investigating the feasibility of enabling them.

Configuring the Email

Next, you need to add the two recommendation scenarios into an email in the Content Studio app. The email contains two product recommendation blocks. The first block uses the recommendation scenario LAST_PURCHASE (QW_LAST_PURCHASE). The second block uses recommendation scenario LAST PURCHASED CROSS (QW_LAST_PURCHASED_CROSS).

Email Editor in Content Studio App

Email Editor in Content Studio App

You can now use your email in an email campaign to send customers personalized recommendations!

Summary

In this blog post, you’ve discovered how to provide personalized recommendations in emails using the following:

  • The Last Purchased Items (Interactions – Optimized) and Often Bought Together (Interactions) algorithms in the Manage Recommendations app.
  • The Content Studio app.
  • An email campaign.

For more information, see What’s New in SAP Cloud Marketing for the 1908 Release.

You can also get more information about the new algorithms, the Manage Recommendations app, the Content Studio app, and Campaigns at https://help.sap.com/mkt.

4 Comments
You must be Logged on to comment or reply to a post.
  • Dear Qiu Wen Li .

    Great blog and great news!

    I did not find the new released recommendation scenarios in my system (1908). Should I raise a ticket or is there anything I should activate in our system?

    BR,
    Guilherme.

  • Hello Guilherme,

    Thanks for your feedback. There is no new recommendation scenarios delivered. To achieve what the blog demonstrates, you can just go ahead to create a new recommendation scenario and a new recommendation model and use the algorithms shown in this blog. What’s new in 1908 are the optimized algorithms.

    Hope this clarify your question!

    Best regards,
    Qiu Wen

  • Hello Qiu Wen Li ,

    thanks for this great blog post!

    It is a nice feature, but I wonder about the definition of “optimized” (aka cacheable) algorithms. My understanding was that the result of an optimized algorithm must depend on the leading/basket items and customer cluster (Target Group) (maybe context parameters as well) only. But this new algorithm depends on the exact contact interactions. So did the definition change or ist this algorithm an exception?

    Referring to the definition “If the result set of an algorithm is unique to an individual user, it is not optimized.” from https://help.sap.com/viewer/13d84c47bb6749a188fd53915c256516/1908.500/en-US/b029c3595ee4457ea8fcb95f3145ed37.html

    This leads to the more generale question: May I create an optimized custom algorithm which depends on contacts (returning different recommendations per contact)?

    Thanks,

    Sascha

    • Hello Sascha,

      Thanks a lot for your feedback and we shall correct our documentation. Yes, you can create an optimized customer algorithms personalized to individual contacts (user to item as we called). We will send you a sample.

      Best regards,
      Qiu Wen