Personalize with Einstein Recommendations
Learning Objectives
After completing this unit, you’ll be able to:
- Identify the steps needed to implement Einstein Recommendations.
- Prepare and upload your recommendations catalog.
- Install the Collect Tracking Code.
Create Personalized Recommendations
Now that you are using personalized content with Einstein’s help, it’s time to turn up the speed on your treadmill and add in Einstein Recommendations, a tool to help you tailor email and web content based on customer behavior and interests. Einstein is not only a brilliant scientist and marketer, he is also a magician. And he is ready to reveal his secrets (well, kind of).
Preparation Is Key
Magic is great when done well, but it requires preparation. Similarly, before you can start sending out magical recommendations, there is some prep work that needs to be done. Let’s start by reviewing the A to Z’s of terms used in Einstein Recommendations.
Term | What It Means | What You Should Know
|
---|---|---|
Anonymous Users (vs. Known Users)
|
A user is labeled anonymous until they become known through an identifier, such as email address.
|
Identification typically occurs when a customer logs into your website or makes a purchase.
|
Attributes or Tags
|
Attributes are actions or content that is tracked and used to make recommendations.
|
There are standard and required attributes, but you can also create custom attributes (brand, color, category).
|
Catalog
|
Catalogs store all the metadata for the info being captured from your website.
|
The catalog helps build a customer’s profile and user affinity.
|
Collect Tracking Code
|
A one-pixel JavaScript snippet used to capture data about user behavior on your website.
|
Data captured from your website, must be tied to a catalog.
|
Contact ID
|
A unique identifier assigned to a customer to identify them from an anonymous user.
|
This ID can never change, so we recommend using a unique key that can be shared between Marketing Cloud and your website.
|
Scenarios
|
Algorithms built within the Customer Intelligence Engine (CIE) to provide personalized recommendations.
|
These are highly tested, prebuilt algorithms that help drive recommendations.
|
SKU/Unique ID
|
Each item in your catalog must have a unique_id (also known as a SKU for products). This catalog field must be unique.
|
It’s important that the value in your catalog can be referenced by your website, so that you can correctly track when a user carts or purchases a unique item.
|
Product Code/Item
|
A product code (or “item”) refers to a class of related unique products.
|
All size variations of an item on the Cloud Kicks website share the same product code, for example. Each variation has its own SKU or Unique ID.
|
How the Magic Works
The trick to all of this is a combination of smart algorithms (a set of instructions which learn over time) that are automatically and constantly being refreshed and fine-tuned for real-time results. Which helps you create highly personalized (read: effective) messages. Like most AI, Einstein Recommendations love data—and lots of it. To get to that data, a tracking code is added to your website to collect customer info from their browser behavior, preferences, ecommerce history, and more.
Then you, a magical marketer, map the data (also known as metadata) you collect from your website to your company’s product info or content in the form of a catalog. Finally, you create content templates for those recommendations.
Clear as a crystal ball? If not, no worries. Let’s walk through a scenario. Our favorite shoe retailer, Cloud Kicks, wants to use Einstein Recommendations. First, the Cloud Kicks team installs the code on their website pages. They then sync a catalog that includes a product photo and the info they want to track about their products.
When their loyal customer Gabrielle Mitchell logs into their website and looks at a pair of running shoes, Einstein records this info and starts to build Gabrielle’s customer profile with her viewing history.
When Yasmin from Cloud Kicks is ready to send an email to Gabrielle, she drags in an email recommendation content block into her email campaign and schedules the send.
When Gabrielle opens the email, she magically sees three different recommended running shoes, including the pair she looked at 2 hours ago. One of the new recommended pairs catches her eye, so she clicks Buy Now.
Gabrielle is happy. Cloud Kicks is happy.
Get Started with Einstein Recommendations
When you are ready, head to Email Recommendations under Einstein.
When you first get started, the user interface takes you through an onscreen step-by-step process to help you implement Einstein Recommendations. You can also go directly to Catalogs and Implementation under the Admin dropdown.
In this section, we want to focus on the two most important elements to get right: the catalog and Collect Tracking Code. Let’s take a closer look at each.
All About Catalogs
Einstein Recommendations use catalogs to understand which assets are available (in stock, published, and so forth) to recommend, as well as what defines those assets (color, price, author, date, and so on). The catalog is where all of your digital assets and the attributes that describe those assets live. Let’s review the three catalog types that can be used in Marketing Cloud.
Catalog Type
|
Description
|
---|---|
Product
|
A catalog of all of your products available for purchase with images of your products and product information. You may have an existing product catalog already that can be used directly or adjusted.
|
Content
|
A catalog of articles, blog posts, videos, or other types of content available for view.
|
Banner
|
A web-specific catalog for image files such as hero graphics, calls to action, or offers.
|
Attributes
Attributes in the catalog drive the rules you can create for your recommendations. So, if you need a unique scenario for your brand, you need to add a custom attribute. For example, if Cloud Kicks wants to recommend a specific type of shoe to a customer based on their search history, Cloud Kicks needs to add a custom attribute of shoe category (athletic, kids, or sandals) to its catalog. What custom attributes are needed for your catalog? It’s important to spend time getting your catalog right, to make sure you have all the needed tags and rules for your content.
Customer Profiles
Your catalog attributes also help build a profile of preferences (called affinities) for each of your customers. Over time, a customer’s profile gets smarter as more data is collected, which leads to better recommendations. Here is an example of how Gabrielle’s affinity and recommendations appears under the Reporting tab and Contacts.
Catalog Import
Next, you need to determine how to upload and sync your content moving forward. Learn more about ways you can import your catalog on the help page, Import a Catalog.
Once you’ve selected your method of upload, head over to Personalization Builder and then to either Email or Web Recommendations. Hover over Admin and select Implementation from the dropdown. Follow the on-screen instructions to set up your import either through upload via streaming updates or through batch upload. Voilà, your catalog is ready to use. But your catalog isn’t the only data Einstein needs, so let’s talk collect code.
Collect Tracking Code
When a visitor lands on your website with the Collect Tracking Code installed, a cookie is dropped with a unique ID and session ID. The cookie tracks the user until it is removed or cleared. Think about how your customers interact with your website. What pages and behaviors would be most helpful for you to track? This data, paired with your catalog, helps build a customer’s profile and affinity—which is why the Collect Tracking Code is so important.
Your company’s collect code can monitor the basics (like browser type, user location, and page URL) and info that might be helpful to you in your marketing efforts (like the category of an article that was read, the brand of a product that was viewed, or if something was left in the cart). Here’s more about these types of data and why you might track them.
Code Types
|
What Does It Do?
|
Why Track It?
|
---|---|---|
Item Detail Views
|
Tracks when users view an item detail page on your website.
|
This is the most important data necessary for recommendations as it logs the content a user is interested in enough to view more information about.
|
Category Views
|
Tracks when users view category and subcategory pages on your website.
|
Category views are necessary for optimal category-page recommendations.
|
In-site Searches
|
Tracks what users are searching for on your website.
|
In-site search details are necessary for delivering search-page recommendations.
|
Cart Activity
|
Tracks when there is an update to an item within a user's cart.
|
This code is sent with all items in the user's cart, along with the quantity and unit price for each item. It helps build individual profiles and is necessary for any cart-abandonment campaigns.Add this code anywhere there is an action to add to
cart.
|
Purchase Activity
|
Tracks when a product has been purchased from a user’s cart.
|
It helps build individual profiles, leads to optimal recommendation quality, and is also necessary for cart-abandonment campaigns.
|
There are many options to customize your Collect Tracking Code to fit your needs. And speaking of tracking, now’s the time to track down your web developer and walk through the code options on the Implementation page together. You can find this page under Admin and Implementation.
Review the list of activities that you want to track and make sure to add the associated code to your webpage. Meet us back here to wrap up your configuration.
Check for Errors
After your code has been added to your website and your catalog is synced, head to the Status tab inside Personalization Builder. Find out if your implementation has been successful or if any errors have occurred that need to be addressed.
Click the link in the Description column to view the error and take steps to fix the issue.
Now that the base configuration is done, you can start the exciting task of creating your customized recommendations. Join us in the next unit to learn more.