This essay originally was published on August 26, 2021, with the email subject line "CT No. 92: Personalization, segmentation, or find-and-replace?" It is the first in a three-part series on personalization. Read parts two and three.
Ah, the promise of digital marketing personalization systems. For years, martech and agency vendors alike have bestowed their customers with the sentence, "We can get the right message to the right customer at the right time." If you're lucky, when you ask them "How do you do that?" your vendor will give the extremely vague answer: "With machine learning" or "With AI."
I still see the "right time" messaging everywhere, maybe because someone discovered that "right message, right customer, right channel, right time" was the right message for my buying persona, literally all the time. Or, more likely, it's because people rely on familiar cliches when describing technology they don't quite understand.
Like its fellow digital debutantes identity resolution and advertising attribution, personalization has been the belle of the media hype ball for years. But neither vendor sales reps nor the average content consumer understands what personalization is because there has never been a unified definition of what it ever meant, for marketing or for content.
Two years ago, my very first post on The Content Technologist attempted to outline definitions for the most common first-party personalization systems in martech. Today I'm going to expand on that a bit, defining some of the personalization tactics we experience as content consumers.
Personalization's promise vs the six types of personalization reality
In the 2010s, at conferences and client presentations, the futurists hyping personalization referenced that one scene in Minority Report when the advertisements assume Tom Cruise would like a Guinness. However, without Philip K. Dick's paranoid vision of crime-reporting mutants and constant surveillance tech, no one has ever successfully explained to me how this completely personalized media ecosystem would ever work with real life data.
In reality, most web-based personalization systems are more accurately described as one or a hybrid of the following:
- Segmentation: a series of user actions give the user one of several content options, where the user is generally unaware of how they're segmented
- Find and replace: Individual words are found and replaced based on user data like name, location, or profession
- Targeting: Users are selected to receive content or promotions based on similarity to other users or location, demographic or behavioral data, usually on a paid+push basis
- Customization: users manually set their preferences either holistically or individually, which they can adjust at any time
- Profiling: a system remembers a logged-in customer's previous interactions or purchases, which can be recalled or referenced on demand
- Recommendations: after a user has viewed or interacted with a piece of content, a system recommends similar content, giving the user the option to opt into the recommendation
Organizations with access to immense amounts of user and content data—Netflix, Spotify, Amazon, TikTok—are able to create hybrid segmentation, profiling, customization, and recommender systems at a massive scale. Because they have access to so many options to serve customers content, this kind of big tech personalization looks effortless.
For the rest of us, we're subject to the lesser capabilities of 2nd and 3rd tier technology, probably deeply failing the 83% of consumers who "expect" personalization. Just kidding, we're not failing anybody. But we have to be more honest and aware of the reality of personalization—and the awkward, if not downright alienating user experiences personalization tech can create.
Otherwise our content future will look more like this:
and less like this: