Two types of relevancy in online marketing

Next Tuesday, I’ll be speaking at the MITX Marketing Tech Series event Get Relevant: The Next Generation of Website Personalization, along with panelists from SiteCore and Unica. If you’re in the Boston area, please do come by.

The promotional blurb for the event is:

A new breed of technology is putting marketers in charge of creating more relevant experiences, experiences that drive higher engagement, sales, conversions and customer loyalty.

Anticipating one of the threads of our discussion, I’d like to ask: are personalization and relevancy the same thing? I contend: not necessarily.

I think there are two types of relevancy in online marketing:

• Expected relevancy: where a user is looking for something in particular, and you’ve implicitly or explicitly promised it to them, and now you have to deliver on those expectations. This is the essence of “message match” that my colleague Anna Talerico is always championing: is a landing page concretely relevant to the ad that generated the click? And yet, while everyone agrees with this in principle, it’s still far from universal in practice.

• Serendipitous relevancy: where auxiliary elements to a user’s primary experience — e.g., advertisements or other kinds of “bubble up” snippets of content in the right or left rail of a page — become more relevant to the user, either in the context of that page or based on a history-based context of the user. Serendipitous relevancy is epitomized by many of the behavioral targeting solutions out there, and there is evidence that it can work quite well… at least if the circumstances are conducive to it.

Most of the time, expected relevancy is highly deterministic. You don’t have to make Bayesian inferences about what a user is interested in. They’ve explicitly told you, through the initial click on a very specific ad tied to a very specific keyword. Still have questions about the segment of that respondent on the landing page? Use a short conversion path to give them direct control over picking what’s most relevant to them (and in the process, self-identify themselves to you for future proactive marketing).

In contrast, serendipitous relevancy is usually a best guess. Based on whatever machine learning or classifying algorithm you’re using, you take a set of data points about a particular user and try to extrapolate who they are (categorically) and what offer might appeal most at that moment.

These are very different types of relevancy.

On one hand, you could say that expected relevancy is reactive: the arrival at relevant content is under the direct control of the user. Serendipitous relevancy, on the other hand, is more proactive: the software is taking the initiative to introduce content that it thinks may be relevant, even if the user didn’t explicitly ask for it.

Here’s a key point though:

People’s tolerance for mistakes in expected relevancy is extremely low. If you don’t give people what they expect in their primary experience flow essentially 100% of the time, they hit the back button right quick.

With serendipitous relevancy, the stakes aren’t nearly as high. You aren’t risking the primary experience of the user. If a behavioral targeting system gets it right 30% of the time, that’s a great achievement. If you miss your mark, there’s only a small chance that you will disappoint the user because they didn’t have great expectations for those auxiliary elements to begin with. In fact, you could argue that when it does get it right, that’s just bonus for you and the user. (That’s why it’s serendipitous!)

Although I believe there are tremendous opportunities for interactions between these two types of relevancy, and the technologies that respectively enable them, I think it can be dangerous to assume that one is the substitute for the other.

Comments

[ixodcom]

expected relevancy is highly deterministic. You don’t have to make Bayesian inferences about what a user is interested in.

[411 WebConnect Canada]

I found very interesting the difference between Expected Relevancy and Serendipitous Relevancy , great insight

411Webconnect Canada

[Chill]

Great post. I like the ideas you raise., I learned so many things on this article.

Thank you!