Recommendation Studio Overview

Blueshift's Recommendation Studio empowers marketers to configure unlimited recommendation schemes to use in messaging templates across emails, push messages, SMS & Display. Recommendations created with the studio can deliver unique & personalized messages for every user in a marketing campaign.

The recommendation scheme is configured in the Recommendation Studio and associated with a messaging template. When a campaign runs, it applies the recommendation steps for every user in the segment to deliver an unique and personalize message with items tailored to the individual user.

Recommendation Studio empowers marketers to build recommendation schemes themselves without the need for IT teams or data scientists. This is an add-on service, please contact support@blueshift.com if you want to use this offering.

In order to use Recommendation Studio, you need an item catalog loaded in the Catalog tab. In addition, you also need to implement real-time click-stream events.

Recommendation Studio works by allowing marketers to create content blocks based on different clickstream or product attributes in the catalog that each user has interacted with. The types of content blocks include:

 

Content type Usage
Based on user’s real-time activity (Events)

Create content based on a user's real time activity across all channels. Content includes items based on behaviors like
abandoned cart, abandoned browse, custom events, abandoned wishlist, price drop, back in stock, favorite items etc...

Any clickstream event with the "Saved products" flag checked will show up in this section, and those items can be used for inclusion in messaging

Based on collaborative filtering (Related items)

Create content based on similar items to user's activity based on other users who have done the same activity. Content includes "people who saw this, also saw that", and "people who bought this, also bought that."

In addition you can create content blocks using product attributes from previous blocks and using them as input to recommendations, for example, top viewed items from the category of the abandoned browse.

Based on user’s affinities (Predictive content) Create content based on machine learning driven models that compute user's affinities to categories (eg: best sellers from user's recent browsed categories), brands (eg: new content from user's favorite brands) and other custom catalog attributes. Predictive content can take user's persistent and transient behavior into account to build recommendations.
Based on site & catalog activity (Item attributes)  

Content based on overall site & category activity. Build content like best sellers, fast selling, trending recently, best value for money across the site and in specific categories.

Eg: New arrivals in accessories, Top trending across site, Most discounted hotels in city.


 

 

 

Default list of recommendations are listed here - Recommendation Types

647b67e-Recommendation_Studio.gif

 

Recommendation schemes created using Recommendation Studio can be used in messaging templates across Email, Push, SMS, Live Content and Syndications. They are available as drop-down options in the respective studios while creating templates.

Recommendation Studio creates a series of blocks numbered 1 through n. Each block may contain one or more products. The output of the recommendation scheme is available as two sets of variables, a flattened set of variables under {{products[0] .... products[n]}} OR as individual blocks.

{{recommendation.block1.products[0] ... recommendation.block1.products[n] }}

  • The nested json has all the product variables like title, url, image, category, brand and custom fields under extended_attributes. See section Personalization for sample usage.

You can optionally test recommendation schemes by attaching different schemes to different templates. For example, create Abandoned Browse with category up-sell scheme vs Abandon Browse with most popular across the site scheme and attach them to different email templates. If the number of blocks and number of products returned are the same, you can code up one template, clone the template and change the recommendation scheme. See section Testing & Optimization for more details.

 

Interested in learning more? Sign up for our Blueshift Fundamentals course.
This course is designed to teach fundamental concepts of Blueshift followed by hands on exercises in a simulated Blueshift instance. 
Was this article helpful?
0 out of 0 found this helpful