Connecting the content model: personalization’s Rosetta Stone

Content pros have long realized the strategic value of a content model. It’s been recognized as a master skill, a central actor in the blobs vs. chunks war, and a starting point for everyone with NPR COPE envy.

For these reasons (and many others), I was geeked out to spend a half-day with Cleve Gibbon and Kate Kenyon this fall at Confab Intensive in Portland in their workshop about content modeling for personalization.

I’ve found (read: stolen) great inspiration from Cleve’s writing and thinking on all things content models. Kate, too, has been “CMS mythbusting” from across the pond for years—a terrific duo to be “stuck with” for three hours on the topic.

Pairing content models with personalization was the stroke of genius. Personalization’s dirty little secret is the fact everyone is talking about it (and talk they do), but hardly anyone is actually doing it well.

The still evolving personalization capabilities within web content management systems (and many point solutions) can leave organizations with a sprawling set of conditional rules that overlay a disorganized content strategy, alongside fuzzy audience insight and personas.

Let’s just say I’m both bullish on the future of contextual experiences and realistic about the current state of affairs.

How does personalization fit together with a content model?

As Cleve explained to the room, “We want to plan, manage, publish and optimize content in a scalable manner.” The key word here is scalable. Kate doubled down on this assertion saying, “You can’t do personalization at scale without a content model.”

That’s a stake in the ground, I’d say. I also couldn’t agree more.

If information architecture is about the website experience, then the content architecture operates in a more decoupled way (see presentation slide below).


Of course, a content model supports a lot more than just personalization. Benefits include better author experiences, more portable content, APIs, and a shared language to use across the origination.

When it comes to extending a content model into personalized content delivery, it’s about connecting it with the data you know about a person. Kate describes this (pictured below) using a combination of three types of data—received, collected and given.



Putting it all together

The workshop used Netflix as an example to demonstrate how to start extending a content model with personalization scenarios. The content model for Netflix (pictured below), includes expected content types such as movies, episodes, series and audience reviews. Each of these content types contains attributes


Looking for personalization opportunities should start with a top user task. The one we used in the workshop was helping a user find a movie they would enjoy watching right now.

Using the above-provided content model as a starting point, we began to brainstorm the types of user data Netflix might have access to. This includes data across the previously mentioned “given, collected and received” categories.

Given data for Netflix may include the customer details such as name and address, as well as additional usage data including length of membership and frequency of use.

Collected data includes the rating data and other survey-based tools Netflix uses to better understand preferences. Received data may include the referring website, location based on IP address, or device currently being used.

By looking for opportunities to connect user data with attributes of the content model, you can start to see the power of personalization when it’s built into the foundation of the content architecture. You can also start to identify gaps in the content model and customer insight gathered based on the opportunities to create more relevant experiences.

Our workshop group brainstormed personalization scenarios such as:

  • Movie recommendations based on time of day: Don’t show Teenage Mutant Ninja Turtle options after the kids go to bed.
  • Movie recommendations based on referring URL: Show the movie of the IMDB profile they just came from.
  • Movie recommendations based on user location: Recommend movies filmed in location (or about the location) of the user’s hometown.

It was clear this approach can help other types of businesses create better planning frameworks around personalization. We use similar approaches in our practice at Connective DX and have also started to extend them to new content types in the world of connected devices and iOT.

More takeaways

It was great to (finally) meet Cleve and Kate in person and to spend a half-day discussing the mechanics of personalization with fellow content professionals. There isn’t enough time spent rolling up our sleeves when it comes to important topics like these.

There were many other tweetable takeaways and sage advice from the workshop. Some of my favorites were:

“No model survives first contact with real content.”

“It is very hard to go back in time and get the content model right.”

“We need to bring people on the journey of creating a content model, not just show them the destination.”

“Content modeling and personalization need to be continuous acts, not fire and forget.”

A big thanks to Kate, Cleve and the Confab organizers for bringing such an important topic to the event. We’ll all be happier with our content management platforms and digital experiences if we can put approaches like these into practice.

About the Author
Jeff Cram

Jeff Cram is Chief Strategy Officer and co-founder of Connective DX (formerly ISITE Design), a digital agency based in Portland, OR and Boston, MA. As the Managing Editor of the CMS Myth, Jeff is passionate about all topics related to content management, digital strategy and experience design.

More articles from Jeff Cram

Leave a Reply
  1. Fields marked with * are required.
  2. We will not publish your email.