Marketers / Analytics / Marketing Cloud

Einstein Content Selection Quick Overview

By Lucy Mazalon

Einstein Content Selection is an email experimentation tool that selects the most engaging content (images) for each subscriber based on the rules you define, subscriber attributes, and what’s been popular in the past.

This tool works in real-time, scanning content in your content “pool” available at the exact moment the subscriber views the email (i.e. not when the email was sent) intelligently taking changes to attributes, others’ engagement, into consideration.

Einstein Content Selection in Marketing Cloud is part of Content Builder, and is one of 50+ Einstein AI-driven products and features. Configure attribute, exclusion, fatigue and variety rules – then Einstein will take care of the rest!

READ MORE: A Guide to 50+ Salesforce Einstein AI Products and Tools

As a premium feature (not available in Basic edition), you can expect significant positive impact. Einstein Content Selection boosted ROI by 18% when used in a promotional email campaign sent by one organization.

How Einstein Content Selection Works

1. Build out a content pool (i.e. your images). This includes asset classes, asset attributes, profile attributes, and fallback assets (we’ll cover an overview of these in the next section). Note that you will have to manually re-add add any assets that are already exist in Content Builder to Einstein Content Selection (hopefully this will be solved in future releases).

Source: Trailhead.

2. Select a data extension, which will provide the profile attributes.

3. Configure attribute rules: This maps asset attributes to profile attributes (i.e. connect the dots between content and your subscribers, which are likely different but mean the same thing in reality).

4. Configure exclusion, fatigue and variety rules: These will be the baseline to guide Einstein in making its choices for which content to serve.

5. Add a Einstein Content Selection block to an email created using Content Builder. If you are adding more than one Content Selection block referencing images from the same asset class, Einstein will know to not show the same image in a single email.

6. Einstein searches your content pool to find the content that subscribers are most likely to click on, based on your rules, subscriber attributes, and what’s been popular in the past.

7. Einstein continues to track subscriber interactions with each piece of content, “learning” what works well, when and with who.

It’s worth noting that Einstein Content Selection dramatically reduces the time it takes to create emails. Not only is the builder drag-and-drop, you save time that would have otherwise been spent agonizingly selecting images and segmenting your Marketing Cloud database to send different versions.

Einstein Content Selection Setup

  • Asset classes: Act like a category for assets, grouping them. This will help with making your analysis more targeted to a specific category of assets. An asset can only belong to one asset class.
  • Asset attributes: Underlying information about the asset (known as “metadata”), e.g. the related product that you sell.
  • Profile attributes: Information about your subscribers, e.g. language, location. This is fetched from the “consumer profile” data extension that you select when setting up Einstein Content Selection.
  • Fallback assets: Assets that will be used if no assets match.

Einstein Content Selection Rules

Configure these rules to ensure that you’re giving business context (i.e. what happens in reality) to Einstein, so it can make the best selections:

  • Attribute rules
  • Exclusion rules
  • Fatigue rules
  • Variety rules
  • Attribute rules: This is where you map asset attributes to profile attributes (i.e. connect the dots between content and your subscribers). The “Must Match” checkbox means that the asset is only for a targeted audience, in other words, only subscribers that have that attribute. This prevents Einstein from straying in its experimentation.
Source: Trailhead.
  • Exclusion rules: Which content shouldn’t be displayed to which subscribers. In other words, hard rules that Einstein must abide by that can’t be inferred from other data Einstein is using for analysis.
  • Fatigue rules: How many times an asset can be used in emails to the specific subscriber – ideally you won’t want to repeat content that’s not sticking. Keep content fresh! 😊 These rules are defined for each asset class, with a fallback fatigue rule to cover undefined classes. Time periods can set to not re-show content:
    • X days after last selection: The last time Einstein selected the content for the subscriber),
    • X days after last click: The last time the subscriber clicked on the content),
    • Selection maximum: How many times a subscriber can see the asset.
Source: Trailhead
  • Variety rules: Switch up content to, basically, give Einstein permission to bend attribute rules. Even if a subscriber is prime to receive one kind of content, you may want them to explore further interests, and so, cross-sell other products.

On the topic of rules, note that Einstein Content Selection is designed for single images. If you want to display more than one image, e.g. a collection of three suggested products, you would risk Einstein showing the same image twice (and managing images for the same product family across multiple asset classes will become a headache). Upvote and follow updates on this idea to see whether this will be changed in a future release.

Einstein Content Selection Dashboard

The dashboard shows the following:

  • No. active assets.
  • No. expired assets (plus, expiring within 7 days, and within 30 days).
  • Asset classes (which act like a category for assets).

Filtered by asset class, you’ll find the following

  • Breakdown of the assets in that class, displayed in a table.
  • Avg. no. selections, i.e. the average number of times assets in the class are selected (within the filtered time frame).
  • A comparison of when Einstein was able to make an informed selection vs when the assets were used as fallback selections, i.e. the “catch-all” option where no matches where possible (displayed in a donut chart).
  • Unique clicks, i.e. the number of subscribers who click on the asset (within the filtered time frame). Take this measurement with a pinch of salt, as some assets are intended to be clickable.
  • Click to open rate, i.e. the % of people that click on the assets out of everyone that’s viewed it (opened the email, within the filtered time frame).

Einstein Content Selection + More Einstein

Einstein Content Selection can be very impactful, especially when paired with other Marketing Cloud Einstein products, such as:

  • Einstein Copy Insights: Text analytics and natural language processing analyzes email subject lines to uncover insights (including the impact of phrases, tones, and punctuation) which you can then use to craft subject lines that drive stronger email open rates. This is the first hurdle you have to jump over in order to gain subscriber engagement – then Einstein Content Selection takes over with optimizing the body of the email.
  • Einstein Content Tagging: Automatically applies searchable tags to image files in your Marketing Cloud account, with help from Google Vision to analyze and process images (which enables you to view asset performance by content tag).
Above: Einstein Copy Insights

See all Marketing Cloud Einstein (and all other Einstein products!) in the guide linked below:

READ MORE: A Guide to 50+ Salesforce Einstein AI Products and Tools

The Author

Lucy Mazalon

Lucy is the Operations Director at Salesforce Ben. She is a 10x certified Marketing Champion and founder of The DRIP.


    stumble guys
    September 19, 2022 3:37 pm
    I frequently check out the videos on the sharing site because the information they contain is so vital.
    April 19, 2024 6:56 pm
    Does Einstein use attributes from the "Consumer Profile" data extension that are NOT mapped to asset attributes? E.g., how would you map an "age" attribute on the consumer profile data extension to assets that are not created to tailor to different age groups, but you still want Einstein to look for those correlations?

Leave a Reply