Off Kilter 205: Personal Insanity.

tl;dr: Here we go again. Again. Sigh.

Every time I watch YouTube, what does its much vaunted ad-personalization engine give me? Ads for diapers I don't need, in languages I don't speak, for body parts I don't have.

For twenty years, marketers have chased the fantasy of the right message to the right person at the right time. What we got is a costly, complex, and bloated tech stack, flattened performance, and, increasingly, strategic irrelevance.

Put simply, this is marketing insanity—doing the same thing over and over while expecting, eventually, to see a different result.

Now the personalization fantasy is being warmed over—this time under the shiny new label of AI: Generative algorithms will write the copy. Machine learning will predict the moment. Artificial intelligence will finally crack the code of human desire.

And, if historical patterns are any guide, marketers will fall for it hook, line, and sinker.

Here's the dangerous reality: marketing personalization, as it's sold and practiced, doesn't work. Not at scale. Not sustainably. Not in a way that justifies the investment or the distraction. This means that adding AI to the equation, rather than resolving the structural problem, will probably just amplify it.

This is because the problem isn't really personalization as such. Instead, it's about definitions. More specifically, the personalization consumers want versus the personalization the Ad/MarTech Industrial Complex sells to marketers.

Consumers want contextual relevance. Don't show me something I already bought. Don't push diapers when I'm shopping for cookware. Don't make me repeat what I just told your chatbot. This isn't magic. It's your customer expressing basic needs.

But that's not what vendors are selling. What vendors mean by "personalization—now with added AI” is something closer to algorithmic sorcery. A promise of real-time behavior-triggered creative, adaptively sequenced across every channel and context, all with the intent of behavioral manipulation, where consumers will have no other choice but to convert.

Unfortunately, 20 years of evidence suggest people don't act this way. Gartner reports that 80% of marketing personalization investments are abandoned within 24 months. McKinsey finds it rarely drives long-term growth. Adobe reports 70% of consumers say digital experiences still don't feel personalized. And Pew finds 79% of Americans are concerned about corporate use of data.

So, why hasn't the personalization zombie died due to its historical underperformance?

Simple, the idea has become hard-coded into the marketer’s psyche, as it repeats across every martech cycle since the early 2000s. CRM would “revolutionize relationships.” CDPs would unify everything. Predictive analytics would automate relevance. Now GenAI will “scale intimacy.” The hype evolves. The logic doesn’t.

However, the result to date has been complexity chaos. Teams have exhausted themselves on the hamster wheel of infinite content. What was sold as automation has become a treadmill that only machines can operate at requisite speed, cost, and scale, which, conveniently, requires even more vendor lock-in.

Here's the uncomfortable reality for the coming boom in AI personalization:

• AI tools depend on training data: Without historical data to learn from, AI systems have nothing to model, predict, or generate, rendering them functionally useless.

• Training data depends on labeling: For AI to make sense of inputs, those inputs must be clearly tagged with meaningful outcomes, so the system knows what "success" looks like.

• Labeling depends on consistent human judgment: Unless humans define and apply labels with clarity and consistency, AI learns from noise rather than truth, resulting in unreliable outputs.

Most brands lack this level of data cleanliness and detailed, well-articulated, well-labeled instructions. What they do have are fragmented systems, unlabeled chaos, and siloed taxonomies.

So what happens? Three things are likely:

• Billions will be spent—again—with little to show for it.

• We’ll confuse volume with impact.

• And we’ll forget Dave Trott’s warning that “shit that arrives at the speed of light is still shit.”

Now, if lack of performance were the only issue, that would be one thing. However, here's where things really break down: Personalization at scale doesn't just fail to perform—it risks actively dismantling brand coherence.

Where brands unify, personalization fragments. A strong brand works like a shared orchestrating system: cohesive, predictable, and symbolically durable. It is a shared social good.

Personalization, by contrast, atomizes that system into a thousand micro-messages and ephemeral signals. Unless the core meaning is already strongly understood, the result will be a brand that's everywhere and nowhere, everything and nothing. Consumers won't know what it stands for—only that it's watching and telling them what it thinks they want to hear.

I'm not sure about you, but I try to avoid hanging out with people who only tell me what they think I want to hear so they can get what they want.

As a result, unless we take action to change the trajectory, there’s a very real potential we’ll crush consumer trust, not with a bang, but with an overwhelming glut of irrelevantly personalized push notifications (I'm looking at you, Uber).

So, what happens if you step off?

Airbnb famously focused on building its brand instead of personalizing its ads. Organic traffic surged. Brand preference strengthened. CAC dropped.

By abandoning the personalization arms race, Airbnb didn't become less relevant; it became more effective. It became more coherent. That's what real brands do.

So, here’s a thought on how we might look at this stuff:

• Treat personalization as a tactical capability, not a strategic imperative.

• Use it where it adds real value and aligns with the consumer's definition, not the tech vendors, such as onboarding, service support, or product education.

• Design for clarity and cohesion

• Be aware of the very real challenge of preparing the site before you start building the house.

Or, put another way:

• Use AI to enhance your brand, not dismantle it.

• Don't believe the Ad/MarTech Industrial Complex hype.

• Ask reasonable questions of zombie ideas before doubling down on them.

• Expend more effort doing the hard work of preparing yourself for the AI future, and less on its shiny baubles.

  ---