Article

Personalization (But Not Creepy)

Paul McAleer

September 3, 2025

An illustration of a (mildly creepy) eye hovering above a phone, with icons representing features like location, history, and user paths.

Years ago I was speaking with a client about rolling out a customized version of their site’s navigation for a certain audience. “We should do something personalized… but not make it creepy.”

There was a chuckle from the other end of the phone call – they instantly got it. And as we developed the business logic for how we’d roll out the navigation, “Don’t make it creepy” was a theme we kept coming back to.

What goes into being creepy, other than referencing an old Radiohead song?

It’s worth talking about what “creepy” means. There’s a balance between our use of technology and wanting it to be anticipatory with not giving up our privacy and feeling surveilled at all times. The ur-example is having an everyday, run-of-the-mill conversation about folding chairs, then opening a social media app to see nothing but ads for folding chairs.

It feels too neat to be coincidental, and that type of action can definitely feel creepy. After all, no permission was explicitly given and the ad has no context or knowledge of what was being discussed (supposedly!) So it ends up feeling like a conversation was being mined for keywords.

Much of this creepiness was borne of early personalization. Amazon was an early experimenter in this by suggesting that customers “like you” bought other things. Shopping for a veggie peeler? Well, a few people who were browsing that same product also happened to buy doorknobs and The Da Vinci Code… so maybe you want to, too?

These crude early days of personalization gave way to more sophisticated algorithms that lined up content with audiences and segmentation and products… but in unexpected contexts. That’s where and how things started to seem weird.

Thankfully, a lot of the software that powers non-creepy personalization have improved with recommendations over time. You might have noticed this too, as shopping and content recommendations have generally, anecdotally, gotten better and less creepy. Regulations on data usage, such as GDPR, have helped wrangle in some of these wild use cases.

When does it make sense to focus on personalization?

So if we have the tools in place, and we have the potential to not be creepy, when should you consider personalization? Here are a few common scenarios.

  • You’re creating a lot of well-written content but it’s not reaching people. Serving up “for you” content is clearly standard practice for social media in 2025. Personalization can open up new distribution channels, owned or not, for this content.
  • People are struggling to interact with your app, website, or product. After a diagnostic (usability testing), you may want to explore solutions that include wayfinding and navigation tailored for a particular audience — personalizing the experience.
  • You’re curious about a truly adaptive experience but aren’t sure where to start. Adaptive experiences are the ultimate expression of personalization: fully-customized, data-driven experiences backed on a foundation of design & content systems. It can be a long road, though, and proving value out with a smaller effort can lead to more options.

What goes into making personalization happen?

Personalization isn’t a strategy, but it can serve one. Here’s what should be in place before you bolster your approach to personalization.

  • A strong taxonomy and governance. Whether it’s in your PIM, CMS, or both, having a consistent and enforced nomenclature and taxonomy will allow any recommendation & personalization engine worth its salt to do its job better. This usually shows up as tags on content, or categories within a product hierarchy. As a bonus: who owns this in your organization? (If the answer is “nobody”, it’s time to start planning for it!)
  • A willingness to experiment. For all the talk of experimentation culture out there, let’s be real: it can be hard to pull it off! A focused team that is willing to make small, meaningful efforts and potentially fail at them is essential. You’ll test, you’ll learn, and you’ll keep going.
  • A full understanding of privacy and opt-in requirements. Any personalization effort should seek consent first, so not only understanding what’s legally required but also what’s ethically best is essential.
  • At least a basic sense of your audiences. You might not have robust segmentation, or even attributes to divvy up your audiences, but at a minimum a working hypothesis of how your users behave and act, coupled with marketing-based personas, can serve as a good starting point.
  • Clear, measurable quant & qual goals. Set your goals before starting anything. Is it about conversion? Engagement? Brand loyalty? Brand awareness? A specific interaction? A specific campaign? A product? How will you know if something works? Critically, how will you not be a creep or a weirdo? Start there, and then determine if personalization could be a tool to help achieve it.

What does the starting line look like?

The first step in building out a personalized experience is a baseline. How does performance look today? What about those goals you’re setting? What’s the competition doing that works well? Where have they failed?

Once that baseline is understood and established, we can flow into defining where it makes sense to experiment and learn. Something like an app home page might not be the best place to start. Another area of your experience, like a Settings screen or an individual Category Landing Page, may be the way to go. Choosing a target that has a good amount of traffic, but isn’t fully critical, helps derisk the effort.

When we’ve got a baseline and a potential target, we can then shift to ideation on an experiment. What’s happening in the customer journey at the moment that would be helpful to address? How can it help us with our goals? And what might it look like? Rolling into a non-creepy concept test with users is a great way to begin collecting data and seeing how the whole thing performs.

From there, there are a lot of potential paths – we might build out a real experiment and launch it to a certain audience, or we might build a library of experiments to decide upon for a later release. In any event, the fact that opportunities open up is a win – it gives you real options to explore.



Over time, personalization as a practice has come a long way to blend information architecture, data, experience, and product into a better end result – something that truly belongs there. Keep it useful, keep it real, don’t be creepy, and you can create something that benefits the people who use your product.