Ecommerce Personalization Examples: What Changed and Why

When people talk about ecommerce personalization, the examples tend to be pretty familiar. Product recommendations, personalized offers, maybe a different homepage banner based on browsing history.

There’s nothing necessarily wrong with any of those. But I think we’re limiting what personalization can do when we treat it primarily as a way to swap content.

The more interesting question is how the experience should change based on what a visitor is trying to do.

Someone coming back to a product they looked at yesterday probably needs something different from someone arriving through Google Shopping. And two people looking at the exact same necklace could be shopping for entirely different reasons. One is buying a gift. The other is buying for themselves.

Why should they automatically get the same experience?

Here are three examples from personalization experiments I’ve worked on. They’re different approaches, but they all started with the same thing: a visitor problem we thought we could do something about.

Three ecommerce personalization examples

1. Give returning visitors a way to pick up where they left off

Imagine spending twenty minutes looking at products, narrowing things down, and finding a few you like. You’re not ready to buy, so you leave.

Then you come back the next day and the site acts like you’ve never been there.

You’re greeted with the latest promotion, a seasonal campaign, or whatever the brand decided belonged on the homepage. Meanwhile, you’re trying to find that product you were looking at yesterday.

We tested a fairly straightforward way to address that. For returning visitors with available product-view history, we showed a Recently Viewed module on the homepage.

We weren’t trying to predict the next thing they might buy. We were using something we already knew to make it easier to continue shopping.

The published case study reported improvements in conversion, average order value, add-to-cart, and product views per visit. That suggests there was value in helping people get back to products they’d considered.

But there’s a nuance here. Just because someone viewed a product doesn’t mean they still want it. Maybe they liked the style. Maybe they were checking the price. Maybe they decided it wasn’t for them.

We see a product view in the analytics either way.

So I wouldn’t treat recently viewed as something that should always take over the homepage. I’d want to understand how often visitors struggle to recover products and whether putting those products back in front of them actually helps.

Read the Recently Viewed case study.

2. Stop assuming a product-ad click means someone wants that product

This one gets at an assumption I’ve seen a lot in ecommerce.

Someone clicks a Google Shopping ad for a diamond ring, so we assume they’re interested in buying that ring. We drop them onto the PDP, show them everything they need to know about it, and hope they buy.

But what if the ring is close, just not quite right?

Maybe the metal is wrong. Maybe the price is too high. Maybe they liked the picture but want to see similar styles. They clicked because it looked interesting, not because they’ve made their decision.

And now they’re on a product page that’s mostly designed to sell them that one product.

We tested showing relevant alternatives to visitors who entered a product detail page through a product listing ad. The alternatives used the landing product as context, giving visitors somewhere useful to go without restarting their search.

Across the all-PLA comparison, the published case study reported a 14.5% increase in AOV, a 9.3% increase in visitor-based conversion, and a 24.3% increase in revenue per visit.

I think the interesting part is what this suggests about the role of the PDP. For some visitors, it’s the place where they make a purchase decision. For others, it’s just another step in figuring out what they want.

Those aren’t the same job.

Of course, the result doesn’t prove that everyone arriving from Google Shopping is still browsing. Some visitors are ready to buy. That’s why I’d want to look at entry behavior, onward navigation, and the whole eligible audience before deciding what deserves more space on the page.

And I’d be careful about measuring only the people who clicked one of the alternatives. Those visitors chose to interact, so their results alone don’t establish the impact of the experience.

Read the PLA Landing Page case study.

3. Sometimes asking is better than guessing

We spend a lot of time trying to infer what visitors want. We look at pages viewed, traffic sources, purchase history, intent signals, all of it.

And I get it. That information can be useful.

But there are times when we could just ask someone a question and do something with the answer.

During the holidays, we had visitors arriving directly on product pages who might have been shopping for gifts. We also had a curated gift-shopping destination, but it wasn’t necessarily easy to find from the PDP.

So we tested a small prompt near the purchase area: “Shopping for a gift?”

The idea wasn’t to interrupt everyone with a giant modal. It was to offer eligible visitors a way to tell us what they needed and give them a relevant next step.

The published case study reported improvements in average order value, add-to-cart, and revenue per visitor. It also described visitors choosing No as spending more time browsing and viewing more products.

That second part is interesting, but I wouldn’t automatically call it a success. More browsing can mean someone is finding useful options. It can also mean they’re still struggling to find what they came for.

The more useful question is whether asking helped visitors make progress and improved business outcomes.

There’s a catch with asking questions, though. You have to be willing to do something with the answer.

If you ask whether I’m shopping for a gift and then send me to the same generic category page everyone else gets, what was the point?

Read the Holiday PDP Prompt case study.

How I’d decide what personalization is worth testing

I wouldn’t look at these three examples and start copying features onto another site.

I’d start with the problem.

Are returning visitors having to repeat searches? Are people arriving through product ads leaving because the landing product doesn’t quite fit? Are visitors struggling to find the right category or getting interrupted when they’re ready to buy?

Those are things you can investigate in analytics, session recordings, usability research, and customer feedback.

Then I’d ask whether there’s a signal we can reasonably use to change the experience. It might be browsing history, where someone entered, what they’re looking at now, or something they’ve explicitly told us.

You don’t need to know their name or have a massive customer profile to start. But you do need to be careful about what you assume from a signal. One click is not a complete explanation of someone’s intent.

I’d also look at whether the change is worth the effort. How many visitors would qualify? Can we actually deliver the experience reliably? What would we measure? And if it works, can we maintain it?

Sometimes the right answer is personalization. Sometimes there’s a basic usability problem that everyone has, and we should just fix it.

If you’re working through which problems are worth addressing, I’ve also written about turning visitor segments into useful experiences.

Measure the experience, not just the feature

One of the traps with personalization is comparing the people who received the experience with everyone else.

If the treatment is shown to returning visitors, they might already behave differently from first-time visitors. If it targets paid product traffic, that traffic may be different from visitors arriving through email.

So if we’re going to claim the personalization worked, I want the comparison to be within the same eligible audience, with people assigned to see the treatment or a control.

For revenue-focused experiences, revenue per visitor is usually where I start. Conversion and AOV help explain what happened. Interactions, product views, and onward browsing can tell us something about behavior, but they aren’t the end goal.

I’ve seen plenty of features get engagement. That doesn’t automatically mean they’re helping the business.

And the opposite can be true. A feature might get relatively little direct interaction but still affect the broader experience. That’s why I care about what happens to everyone eligible for the test, not just the people who clicked.

I go into that more in how to measure ecommerce personalization.

The opportunity is bigger than swapping content

What I like about these examples is that they don’t depend on knowing everything about a visitor.

We used what someone had viewed before, where they entered, or what they told us they needed. That was enough to create a different experience worth testing.

That doesn’t mean customer profiles aren’t valuable. They are. But I don’t think we should wait until we’ve identified someone and built a perfect profile before we try to help them.

I keep coming back to the same question: why are we showing everyone the same experience when we have reasonable evidence that they’re trying to do different things?

And if we can’t explain how a personalization is supposed to help the visitor, maybe it’s not the next thing we should build.