Challenge	Automate product recommendations, increase the relevance of offers on the site, and improve interaction with loyal customers
Solution	Implementing personalized recommendations on different types of website pages
Implementation time	3 months (including technical integration and customization)
Resources	Yespo integration team and customer success manager, SEZON marketer
Results	+17% time spent interacting with the site +30.5% products added to the cart +244.6% products added to favorites +389.2% page views 42.9% “Similar Products” block CTR

SEZON always knew that customer convenience is key to driving repeat purchases. The brand team wanted to automate part of the interaction, make it more accurate, and better tailor it to the interests of each customer.

The solution was implementing product recommendations on the website—a tool designed to showcase relevant products at various stages of user engagement. We collaborated with the Yespo team to develop and refine the system, gradually testing different approaches and configurations.

In this case study, we’ll briefly outline where we started, how we worked with product recommendations, the results we achieved, and what the company plans to do next.

About the Project

SEZON is a Ukrainian brand that has been selling shoes for the whole family for over 10 years. What began as a single store in Chernivtsi has grown into a national chain with more than 40 locations across major Ukrainian cities. During this time, customers have purchased over 1.6 million pairs of shoes. SEZON continues to expand its network of physical stores while simultaneously strengthening its online presence.

The brand’s core audience consists of modern, active women aged 35-54, who are looking for comfortable, stylish, and high-quality shoes for everyday wear, travel, and special occasions.

Challenge

The SEZON team aimed to move away from manual product selection in recommendation blocks and transition to an automated system that adapts to each user’s actions. The primary goal was to make customer interactions more accurate and dynamic, enabling the system to better respond to the interests of loyal customers.

With the growth of online sales, the team decided to test product recommendations as the first tool for automated personalization. These were implemented on the website in collaboration with Yespo CDP.

As SEZON had not previously used other CDPs, it was crucial to find a solution that would be both flexible and easy to understand from the outset. The focus was on achieving the following objectives:

• automating product recommendations on the website, instead of the need for manual product selection;
• enhancing personalization in interactions with loyal customers;
• increasing communication effectiveness through more relevant offers.

Solution

Customizing Recommendations on the Website

Recommendation blocks simplify navigation and create a personalized experience for each visitor. They display products that a customer may be interested in, even if they weren’t specifically searching for anything. This helps maintain their attention, encourages them to explore the site further, and ultimately drives sales.

The integration of recommendations involved several key steps:

1. Connecting the Professional tariff to access advanced personalization features.
2. Setting up web tracking to record users’ actions on the site, such as product views, category visits, and purchases. This data is then transferred to Yespo and serves as one of the core foundations for forming recommendations.
3. Defining the pages where recommendation blocks would appear and coordinating this with Yespo’s Customer Success Manager.
4. Selecting algorithms for each block based on the page’s goals, such as attention retention, upselling, or offering related products.

All recommendations are generated automatically, based on the transformer model. It analyzes the user’s purchase history and actions, examines the sequence of their steps, preferences by category, price, and time.

This data is converted into numerical vectors, which the model compares to predict the most likely interests. As a result, the system tailors personalized recommendations for each user, without any extra action from the team.

By the way, transformer models show 71% higher efficiency than traditional models. You can learn more about how it works from our article.

Recommendations on the Homepage

The homepage is often the first point of contact for visitors. Since users may arrive without a clear intent, the primary goal of the recommendation block is to capture their interest and encourage them to explore further.

At this stage, we introduced the display of popular products—those that are most frequently viewed or purchased by other users. This creates a social proof effect: “If others like it, maybe it will suit me too.” This approach fosters trust from the outset and motivates users to browse the catalog.

Recommendations in Categories

On category pages, visitors are generally more purchase-oriented. They are browsing a specific group of products but have not yet made a final decision. To assist them at this stage, the team added a block showcasing bestsellers—models that are most frequently purchased by other customers.

These recommendations help users quickly identify popular and proven options. For users, they act as a reference point and guide, highlighting what’s worth considering first.

Recommendations on Product Pages

To maximize the potential of the product page, we introduced two types of recommendations: related products and alternatives.

Related products help increase the average order value (AOV) by encouraging users to add items that are logically connected to the one they are viewing. Alternatives, on the other hand, keep users engaged if the main item doesn’t fully meet their expectations. This reduces the risk of losing a potential customer and increases the likelihood of a purchase.

Products that are Often Bought Together

This block shows what is usually bought together with the selected item. For example, for a pair of shoes, the user will see shoe care products or insoles. The algorithm is built to suggest add-ons from a relevant category, taking into account the previous purchases of other users.

 

Similar Products

Similar models, based on color, material, and style, are displayed next to the main product. This allows users to compare options, choose the best fit, or explore more alternatives before making a decision.

Recommendations on Page 404

A 404 page typically indicates that something went wrong and the user didn’t find what they were looking for. However, even in such situations, it’s possible to maintain their interest. We decided to leverage this moment by adding a block with popular products to the 404 page.

This approach helps to keep the customer’s journey on track by offering relevant options based on what’s in demand among other users. In some cases, this is exactly what encourages them to continue browsing or even make a purchase.

Results

Within the first two months following the implementation of product recommendations, SEZON observed notable changes in user behavior on the site. Key metrics began to show significant growth:

• Website interaction time increased by 17%.
• Items added to the cart rose by 30.5%.
• Additions to favorites surged by 244.6%.
• Page views per session grew by 389.2%.

The “Similar Products” block, in particular, demonstrated exceptionally high engagement, with a click-through rate (CTR) of 42.9%, the highest among all recommendation types.

Plans

SEZON does not plan to stop at implementing product recommendations. The team is gradually moving to a more comprehensive work with the customer base through personalized email campaigns, flexible communication workflows, and deeper analysis of interactions.

Key priorities moving forward include testing new formats, increasing user engagement, and enhancing customer retention:

Conclusion

The introduction of product recommendations provided SEZON with an opportunity to test automation and personalization on its website. Even with a basic setup, the brand observed notable changes in user behavior, including increases in cart additions, favorites, page views, and time spent interacting with the site.

This case study demonstrates that personalized recommendation blocks can significantly influence user interactions when placed strategically, configured properly, and regularly analyzed.

If you’re interested in exploring personalized recommendations on your website, simply fill out the form below. We’ll discuss your business needs, demonstrate the possibilities, and recommend the best solution tailored to your startup.

Get professional expertise