Read time 5 minutes
Better, Smarter, Faster: How AI is Transforming CDPs
Why it matters:
Marketers who read this post will walk away with an understanding of why most personalization programs miss the mark, and a clear framework for fixing it. They will see why even sophisticated programs end up confidently wrong about the customers they are trying to reach. And they will leave with a four-rung ladder to climb from guesswork to a genuinely personalized customer experience.
Key takeaways:
Watch Optimove's Ryan Johnson and Alana Yentis introduce the Personalization Ladder at Optimove Connect 2026:
Ryland Johnson, who leads product for Optimove Personalize, does not love cats.
Yet, he opened this Connect 2026 session with a story about cat food.
He bought a bag of it. He does not own a cat. He was looking after someone else's. The next week, the retailer started recommending cat toys. The week after, urine pads. The week after that, joint oil for cats.
The retailer had built a whole life for an imaginary geriatric cat from a single grain-free purchase. And the most interesting thing about it, Ryland said, was not that the retailer was wrong. It was that the retailer was confidently wrong.
That phrase, “confidently wrong” is the spine of the session. Most personalization in marketing today is confidently wrong about most of the people it is trying to personalize for. It does not lack data. It lacks the discipline to earn its conclusions one rung at a time.
Roughly 70% of businesses never get past the first rung of personalization... personas such as fictitious Sally, age 40, two kids, two dogs, likes 80s pop. We have all sat in the meeting where Sally was described. We have all nodded along. And we have all quietly known that Sally does not exist.
The hope is that the email blast is wide enough that people like Sally receive it. That is a workable strategy. It is also a confidently wrong one.
The businesses that move past personas usually leap to whatever sounds most advanced... business rules, then machine learning, often without proving anything along the way. The result is a familiar one: more sophisticated personalization that is, on a different axis, confidently wrong about more people, faster.
What this Connect session laid out is a different approach. Not jumping to the top of the ladder. Earning each rung.
Each one earns the right to climb to the next.
Rung 1: Hot right now. Start with what is actually provable. Show every user the content that is genuinely popular, right now, based on actual on-site activity. No assumptions about the individual yet. No invented personas. Just the truth of what is working, served to everyone. It is the most honest starting position, and most teams skip past it because it feels too simple.
Rung 2: Reactive personalization. Once popular content has proven itself, earn the right to get specific. React to what the individual user actually did. Do not react to what was assumed about them. This is also where business rules belong, layered on safely, without creating the "bus factor" Ryland warned about (the risk of business logic so complex that only one or two people on the team understand it).
Rung 3: Predictive personalization. With reactivity proven, move to anticipating what a user will want before they ask. Look at users who behave similarly and surface what they engaged with. This is where machine learning earns its place, not as a shortcut past the lower rungs, but as the natural next step for a team that has already proven the rungs below.
Rung 4: Cross-channel personalization. Take the intelligence from the three rungs below and apply it everywhere the user touches the brand. Web, banner, search, email, SMS, push. The recommendations are not identical across channels, but they are consistent, relevant, and built on the same foundation.
Alana Yentis, who leads customer success for Optimove Personalize, made the case throughout the session that the ladder is not a feature list to choose from. It is a sequence to follow.
Skip a rung, and you build personalization on assumptions you have not yet earned. Start at rung four with no foundation underneath, and the recommendations look advanced but rest on the same imaginary geriatric cat. Start at rung one and prove it, and every subsequent rung gets stronger because the data underneath is real.
The Netflix recommendation prize is the cautionary tale. Ten years ago, Netflix offered a million dollars for the best recommendation algorithm. Someone won. Netflix did not implement most of the winning approach. The recommendations were too good. Users saw the same suggestions every time they logged in, viewing time stayed flat, and login frequency quietly dropped. The metrics looked fine.
The system was confidently wrong, and the metrics agreed. The Netflix viewers got bored from monotony of seeing the same recommendations repeatedly.
Even at the top of the ladder, confidence has to be earned.
Confidence in personalization is not a starting point. It is something a program builds, one earned rung at a time. The marketers who win do not jump. They climb.
Learn more about Personalization across the customer journey.
For more insights, contact us to request a demo.
Whitepaper: How AI is Transforming CDPs
CDP Institute’s David Raab shares what business leaders should start thinking about now to take advantage of next-generation CDPs.
Rob Wyse is Senior Director of Communications at Optimove. As a communications consultant, he has been influential in changing public opinion and policy to drive market opportunity. Example issues he has worked on include climate change, healthcare reform, homeland security, cloud transformation, AI, and other timely issues.
FAQ
What does “confidently wrong personalization” mean?
It’s when a brand makes strong personalization decisions based on weak or unearned assumptions—like building a full customer profile from one purchase signal.
Why are personas a weak foundation for personalization?
Personas can be useful for messaging direction, but they often rely on fictionalized traits and assumptions that don’t reflect real individual behavior.
What’s the best place to start improving personalization?
Start with what’s provable: show everyone what’s genuinely popular right now based on real on-site activity—then build from there.
Where do business rules and machine learning fit on the ladder?
Business rules belong in reactive personalization once you’re responding to what users actually did. Machine learning earns its place at the predictive rung, after the lower rungs are proven.
Why can even “better” recommendations hurt performance?
If recommendations become too repetitive, users can get bored—even if the suggestions are accurate—leading to flat viewing time and fewer logins.
