I was scrolling through my shows and couldn’t find Billions. Not because it wasn’t there. Because I was looking for it. I kept scanning past it. Paused. Scrolled back. Still missed it.
Until I realized what changed. The cover image. Instead of the cast I associate with the show, it was a different character. Still from the show. Still relevant. But not what I use to recognize it.
It didn’t change the show. It changed how I find it. And that’s the point.
This is where personalization breaks.
Most platforms assume that changing what you show will make something more relevant. Sometimes it does. But sometimes it interferes with something more important: recognition.
People don’t navigate interfaces by reading. They scan. They rely on familiar visuals, repeated patterns, and quick identification. You don’t search for what you’ve already chosen. You look for what you recognize.
So when that visual anchor changes, even slightly, the cost isn’t confusion. It’s friction. Not enough to stop you completely. Just enough to slow you down. Make you second guess. Make you miss it once. Then twice. That’s enough to break the experience.
The Misunderstanding
This wasn’t random. It was likely an attempt at personalization, or representation, or both. That’s not the issue. The issue is the system misunderstood the job of that surface.
The cover image isn’t there to experiment. It isn’t there to optimize identity. It isn’t even there to drive engagement. It exists to do one thing: help you recognize what you’re looking for immediately. When that function is disrupted, everything downstream gets harder.
The Behavioral Layer
Not every surface should be personalized. Not every interaction should be optimized. Not every variable should be tested. Because different parts of a system do different jobs — and optimizing one can break another.
- Some surfaces are for discovery, persuasion, exploration
- Others are for retrieval
And retrieval depends on stability. Consistency. Familiarity. When you personalize a retrieval layer, you risk breaking recognition. And once recognition breaks, navigation slows.
The Broader Pattern
This doesn’t just happen in streaming. It shows up everywhere.
- Product thumbnails that change too often
- Navigation labels that get “optimized” into ambiguity
- Landing pages that shift structure between visits
- Interfaces that prioritize novelty over recognition
I saw this working with a fashion brand. The creative team wanted to display clothes in editorial, high-concept styling — interesting for awareness, compelling for brand identity. But the product pages were pulling from the same imagery. Someone arriving to buy a specific jacket wasn’t exploring. They were retrieving. The editorial images weren’t wrong. They were in the wrong place.
The Strategic Takeaway
Separate the decision layer from the recognition layer.
- Personalize where people are deciding: recommendations, offers, messaging
- Stabilize where people are navigating: labels, structure, visual anchors
Search results, navigation, and “continue watching” modules should stay highly stable. Hero banners, carousels, and promotional tiles can carry the personalization. Because those are not the same job.
The Reframe
Personalization is often treated as universally good. More relevance. More engagement. Better outcomes. But relevance isn’t always about what you show. Sometimes it’s about how easily someone can find what they already decided they want.
If you change how something is recognized, you change how it’s used. And not every system can afford to trade recognition for relevance.