Personalized content recommendations represents the next generation of email marketing.  The primary purpose of email marketing is to drive traffic to a web site.  The cycle is completed when a site visitor’s own page view history determines what type of content he receives in future emails.

Maropost Marketing Cloud implements state-of-the-art machine learning that works behind the scenes tracking each contact’s own web site page view history.  Based on this stored knowledge, marketers create Rule Sets for personalized content that is automatically inserted into each contact’s own email at send time.

Content Recommendations requires the following elements:

  1. HTML meta tags that the web developer inserts into each web page.
  2. The Maropost web spider that crawls the web site, cataloging each page in the Web Content Library
  3. The Maropost web tracking script that the web developer inserts into each web page.  This script builds the entire page view history of each contact.
  4. A Rule Set editor that the marketer uses to define the specific types of content that the Recommendation Engine should select when determining “best match” for each contact.
  5. A content feed that the marketer inserts into the content of the email. At run time when the emails are created the Recommendation Engine will automatically insert  references to pages on the web site that best match each contact’s own page view history.

The Web Site Library

The Web Site Library is a highly-indexed catalog of a web site.  The Web Site Library contains meta data about every page on the site.  The Web Site Library is the source of content that will ultimately be displayed as personalized content within every recipient’s own email.  

The Maropost web crawler, or web spider is a specialized Internet bot that visits every page of the designated web site, making an entry in the library for each page visited.  Web masters can restrict what portions of the web site that the spider can view or not view.  The rules are defined in a robot.txt file that is stored on the web site.

The purpose of the web spider is to build the web site library by cataloging each web page it visits.  The spider does not, however, build an exact reproduction of each page of the web site.  The web developer tells the web spider how to catalog the page by inserting META tags into the web page’s HTML code.  META tags are not displayed when the site visitor views the web page in his browser.  Rather they are used to pass instructions to a web browser or a web spider.  The three standard META tags that the web developer uses to catalog a web page to a web spider are Keywords, Description, and Author.

Open Graph META tags are a new type of META tags that Facebook introduced  to facilitate the “sharing” of web pages within Facebook posts.  Open Graph META tags are defined by the prefix “og:”. For example, <meta property=”og:titlecontent=”Cinnabon’s Cinnabun Copycat Recipe” /> describes the web page title containing a recipe for a cinnamon bun.  Most modern web spiders have the ability to read Open Graph META tags since they are so widely used in web sites.  Common Open Graph META tags include og:title, og:description, og:image, og:type, and og:url.

The Maropost web spider will also read custom META tags.  For example, maro:keywords is a custom META tag that enables the web developer to provide keywords describing the web page to Maropost’s own web spider as it catalogs the page in the Maropost Web Site Catalog.  The maro:keywords META tag is a required tag because it is the primary way that the Maropost web spider categorizes the web pages that a site visitor has viewed,  as well as pages having similar groupings of keywords that he has not yet viewed.

As the web spider crawls across every page of a web site that it has access to, it builds a catalog of the web site in which each web page is defined by the meta tags — and especially the keywords — that the web developer has inserted into the page’s HTML code.  The Web Site Library becomes more than just individual records of META tags and keywords.  It becomes a collection of keywords, how they tend to be grouped together, and the frequencies of each grouping.

Web spiders regularly crawl a web site in order to keep the Web Site Library as up to date as possible since web developers are constantly adding new content and new products to the web site.

The Site Visitor’s Web Profile

The Maropost web tracking script that is embedded in every web page records the collection of META tags and keywords of each page that the visitor has viewed.  Over time, the site visitor’s Web Profile becomes defined by the META tags of the pages he has viewed and the frequency of each META tag.  More importantly, it is the combination of META tags — in particular the keywords and the combination of keywords passed in by the maro:keywords META tag, and the frequency of each combination that provides the most valuable insight in defining the site visitor’s Web Profile.

The combination of keywords, and the frequency of the combinations from the site visitor’s page view history establishes the types of content he has viewed most frequently on the site, and is most likely to view in the future.

The Recommendation Engine

The Recommendation Engine defines the entries in the Web Site Library that best match each site visitor’s Web Profile.  The Recommendation Engine operates in near-real time. Even though web sites remain fairly static, a site visitor’s Web Profile is constantly being updated with each page that the visitor views.

The marketer has the ability to tell the Recommendation Engine which combination of META tags and their values are to be used when defining the entries in the Web Site Library that best fit the site visitor’s Web Profile.  The keywords recorded by the maro:keywords META tag and the standard HTML keywords META tag are the primary means by which the marketer instructs the Recommendation Engine to find content that best matches each site visitor’s Web Profile.

In addition to keywords, the Recommendation Engine includes other content matching rules based on products that a contact has previously purchased, landing pages he has viewed in response to an email campaign, content that was published within a particular date range, or even localized content.

Each META tag and it’s value has a weighting factor that enables the marketer to instruct the Recommendation Engine which META tag and value has more of a relative influence over the other tags and their values.  The unique combination of META tags and their values together with their respective weighting factors is a recommendation rule.  By default, only content whose date defined by the maro:expiry META tag.has not yet passed and content whose pages where that META tag is not used.. The ignore expiry option instructs the Recommendation Engine to consider all content, including those whose expiry date has passed.

Marketers create and save multiple recommendation rules for re-use depending upon the type of recommended content they wish to ultimately appear within any given email.

The Recommendation Engine starts with each site visitor’s Web Profile and finds the content in the Web Site Library that best matches according to the recommendation rule that the marketer specified.

Usually, the marketer desires to display several blocks of recommended content within a personalized email.  The recommendation rule specifies both the minimum number and the maximum number of entries in the Web Site Library to recommend for each site visitor.  The Recommendation Engine returns the entries in the Web Site that best fit an individual site visitor’s Web Profile in the form of an array of entries within a structured data file.  The structured data file that is unique for each site visitor based on the marketer-defined recommendation rule and the site visitor’s Web Profile is a personalized content feed. The personalized content feed includes elements that will appear in the body of the email.  Such elements include the title of the content (not necessarily exactly the title of the web page), a description of the content, an image, and a link to the web site page.

The Personalization Engine

Maropost supports static content feeds, that is, content feeds whose elements are the same for all recipients of an email campaign.  The Personalization Engine is a specialized application of creating emails from content feeds.  It reads personalized content feeds produced by the Recommendation Engine for each recipient of an email campaign, and injects its content at runtime.

Elements of a static content feed are accessed within an email using Liquid scripting syntax.  Elements of a personalized content feed are accessed in an identical manner.  Other than being unique to each email recipient, elements of a personalized content feed are rendered in identical manner as static content feeds.  They can be iteratively looped, links within the personalized content feed are tracked, the content is displayed in the web-hosted version of the email (along with the tracked links), and the personalized content feed can be referenced by a dynamic content block.