Personalization on Steroids: AI Recommendations in iPresso

Most personalization systems are simple drawing machines with an AI tag attached.

Inserting a name into an email isn’t personalization. Showing a customer a recently viewed product isn’t personalization. It’s simply operational memory that adds nothing to the customer experience.

Real change happens when marketing automation systems integrate with AI recommendation machines, such as Vertex from Google.

This isn’t just another add-on, it’s a change of engine. Switching fromI remember what you clicked to the modeI know what you’re going to do in a moment.

If you follow what AI does with its models and how search engines are changing, you see a common denominator.

Exactly the same thing can happen in your e-commerce if you put AI models to work.

Why does e-commerce need to start working like Google Discover?

Think about why people spend hours on the Discover feed or on TikTok.

They don’t really look for anything there – they just get it.

The system knows better what will interest them.

For years, e-commerce was a library; the user had to know what they were looking for to find it.

With AI-powered recommendations, your store will become a feed.

Signal Reweighting: The End of the CTR-Only Era

In the old recommendation systems, only the click mattered.

If something clicked well, the system displayed it to everyone.

The result? Promoting clickbait products that generated either returns or frustration.

In the model that iPresso implements thanks to Vertex AI, the weight of the signals is distributed completely differently.

What counts is the satisfaction after clicking.

If a user clicks on a recommended product, spends 3 minutes on the website, reads the specifications and adds it to the cart – the model receives a signalThat was it.

If he enters and leaves after 2 seconds – the model receives informationI was wrong, it was an empty click.

The algorithm learns not how to attract attention, but how to satisfy a need.

Thematic stability – the key to trust

Google heavily rewards topical authority. If your site is about everything, it’s about nothing.

It works identically in product recommendations.

AI learns specialization user in a given session.

If someone came for hiking boots, the system will not suddenly suggest a suit – even if there is a promotion for it.

The system makes sure not to throw the customer out of flow.

Statistics

Before we get into the technical details, let’s see what the analyses from the last few months say.

• 71% of consumers expects personalized interaction, but more importantly –75% declare they will immediately switch to the competition if the experience is not tailored to their needs (McKinsey – Growth, Marketing & Sales Insights 2026).
• AI-powered personalization generates10% to 15% revenue growth(even 25% for market leaders). Product recommendations are already responsible for31% of total turnover in sessions in which the client used them (Involve.me, Marketing Personalization Statistics 2026).

Recommendation models

In iPresso, you have access to specific strategies that Vertex AI optimizes in real time.

1. Recommended for You

This is your equivalent of the TikTok algorithm.

The model analyzes the entire user history in the CDP.

Knowing that the customer bought a camera three months ago, was looking for a lens a month ago, and was reading a guide on night photography yesterday, the system will display a tripod.

Not because you have it on sale. Because it’s the logical next step in his passion.

2. Others You May Like

This model focuses on the “here and now.”

If a user enters the page and starts browsing for slim-fit white shirts, the system adjusts the frame.

Even if this user usually buys t-shirts.

The model understands that the intention is different in this particular session. It predicts the next step within the current context.

3. Frequently Bought Together

AI analyzes millions of baskets and finds correlations.

It may turn out that people who buy a specific model of headphones almost always also buy a specific case that is not in the same category.

The system will detect this itself and start promoting this set.

4. Recently Viewed

Simply displaying recently viewed products is boring.

Vertex AI in iPresso sorts these products by their likelihood of returning.

If a customer looked at 5 pairs of shoes, but spent 2 minutes on one and returned to the size chart twice, those shoes will be at the top of the frame.

Summary

The era of “pouring water” in marketing is over, because AI (both on Google’s side and on the user’s side) is detecting missing values ​​more and more quickly.

If your store still operates on the principle of “let’s show him anything, maybe he’ll buy,” you’re losing money on every session.

Implementing Vertex AI recommendations in iPresso is not a two-year project.

It’s a decision about whether you want to continue to manually control traffic or whether you prefer to rely on an engine that learns faster than any team of marketers.

The technology is here. The data is with you.

Just connect them.

If you want to talk about how to specifically connect your events to Vertex AI and which models to launch at the start –let me know. 

FAQ

What is Vertex AI-powered recommendation engine?

The Vertex AI-powered recommendation engine is an advanced machine learning system that analyzes user behavioral data and product metadata to predict future purchase intentions. Unlike simple rule-based systems, these models use neural networks to rank products in real time, tailoring offers to individual customer profiles (1:1 personalization).

How does iPresso integrate with Google Cloud Vertex AI?

iPresso integrates with Vertex AI via direct transfer of behavioral events from the Customer Data Platform (CDP). User interaction data is sent to Google Cloud models, where it is processed by ranking algorithms. Results (recommendations) are returned to iPresso and displayed to users as recommendation frames on websites, in mobile apps, or in open-ended emails.

What are the benefits of using AI in e-commerce personalization?

The main benefits include increased conversion rate (CR), increased average order value (AOV), and improved customer retention (LTV). Thanks to better product relevance, users spend more time on the site (Dwell Time) and are less likely to abandon their carts. AI also allows for the automation of processes that previously required manual rule setting by marketers.

What is “click satisfaction” in recommendation models?

Post-click satisfaction is a qualitative metric used by advanced AI models to assess the real value of recommendations. Instead of measuring only clicks (CTR), the system analyzes user behavior after landing on a product page—session duration, interaction with the description, and purchase completion. This allows the models to promote products that actually meet user needs, not just attract attention.