Personalization is good, but if things get too personalized, you’re just living in the Matrix
Personalization is usually a good thing. For example, lots of testing in many markets over many years has shown that an email that begins “Dear Greg” is going to do better than an email that begins in some generic way. Assuming, of course, your name is Greg.
That level of personalization is easily tested.
But as you go deeper into personalization, new issues arise.
- Too much personalization might make users uncomfortable, and it might raise concerns about privacy.
- Personalization depends on having the right data – and using it properly. Are you sure the data you’re collecting is fit to purpose, and that your algorithm is tuned appropriately?
- How does your algorithmic personalization interact with explicit user preferences? For example, a reader says he wants financial information, but your algorithm says he prefers sports. Which one gets the upper hand? You have to wonder if you should listen to what users say versus observing how they behave.
- To what extent should you disclose how your personalization engine works?
Beyond these questions, you have to wonder if users actually want it. Too much personalization might lead a person to believe he’s missing out on something important, or is living in a bubble.
In fact, if you had perfect personalization, that seems to be getting into the realm of solipsism, which is the idea that you’re the only mind that exists.
Readers might prefer content that’s curated by a trusted professional over content that’s curated by a machine. The trusted professional knows more about the topic than my browsing history.
Also, personalization can get icky.
As a silly example, it’s one thing for Spotify to say that people who like Jethro Tull might also like Fairport Convention. It’s another thing for Spotify to notice that I was chatting up the redhead at the office party and to tell me what kind of music she likes.
A final thing to consider is a strange side-effect of personalization that Bob Hoffman has discussed in the context of advertising.
Part of the effect of an ad is that I know other people are seeing it as well. If I see James Bond driving an Aston Martin, and I want to be as cool as James Bond, I might buy an Aston Martin.
But the algorithm knows that if I won the lottery tomorrow and had money coming out my ears, I would never spend the money on a fancy car. That’s just not me.
So in my version of the James Bond movie, I see James Bond driving a Toyota Highlander, and I know that you’re seeing him driving something completely different. The effect is completely lost. I can’t be “as cool as James Bond” by driving what he’s driving, because everybody sees a different thing.
I’m not trying to dissuade anybody from pursuing personalization. It’s a proven effective strategy. But the personalization tools we have now, and will have in the future, might force us to re-evaluate.
Links
