English
Welcome to the next episode of What AI Actually Does in Your Store. Last week we looked at how search and recommenders learn from your shoppers. This week: does your store actually personalize, or just segment?
This series is about what AI does in online stores, based on what e-commerce managers ask, worry about, and get wrong.
E04: Do you personalize or just segment?
Many online stores that claim they personalize content individually to each shopper don’t really do it.
Showing the same winter jackets to Dutch shoppers in November is not personalization. Same for pushing bestsellers to new visitors and new arrivals to the returning ones.
That’s just segmentation.
Such a system doesn’t really know anyone.
Real personalization
A shopper opens a product page and your store shows them a row of product recommendations below the header. Who picks the items?
It’s the recommender that combines which products shoppers buy together, which products complement each other in the catalog, what the shopper is doing right now, what it has learned about them from past visits, and what’s in stock.
A new arrivals widget in a fashion store shows what that means concretely:
- pull new items from the categories they usually shop
- filter them by their gender and size
- weight toward their usual price range
That’s the real personalization. AI making a decision in real time, for that specific person.
Two types of personalization
The recommender draws from two layers of data.
Short-term intent: what the shopper is doing right now. This session’s clicks, the query they just typed, the product they’re looking at. It catches a returning customer who’s suddenly shopping for someone else.
Long-term preferences: what we’ve learned over many visits. Their categories, brands, sizes, price levels, and style. It catches a shopper who keeps coming back to women’s size 38, no matter what they search for.
The majority of shoppers don’t give you much of a long-term layer. That includes new visitors with no history, shoppers who declined cookie consent, or anyone browsing in incognito mode.
And that’s fine.
But it’s exactly why a recommender system must work well for people it doesn’t know yet. Personalization is a layer on top, not the core engine.
It pays off where it matters most: returning, high-intent customers, the ones who drive most of your repeat revenue.
Quick self-check
If you turned off your personalization tomorrow, would your homepage look different? Would any individual shopper see different products than last week?
If the answer is mostly no, you’re only segmenting. But the technology to do more has been available for a while.
Next week: Behavioral AI infers what shoppers want from what they do, but some shoppers don’t know where to start.
