How to Increase Sales in the AI Era: Consumer Electronics Retailer Experience

Foxtrot is the first Ukrainian electronics and home appliances retail chain. Today, they are one of the country’s largest omnichannel retailers, which constantly develops personalization of customer experience for 13 million customers and implements new tools and mechanics.

Over 11 years ago, Foxtrot strengthened this direction by choosing Yespo as a reliable marketing partner with whom it’s convenient to grow, and every idea receives rapid implementation.

During the recent expo event of Association of Retailers of Ukraine, Oleg Nikolsky, CMO of “Foxtrot,” and Oleg Lesov, CPO of Yespo, shared their experience of successful collaboration between the retailer and the customer data platform. They explained how innovations in personalization and product recommendations contributed to sales growth and helped in retaining customers and maintaining loyalty.

In this article, we’ve gathered the key points, from the first email campaigns to advanced AI solutions and future plans.

How the Collaboration Between Foxtrot and Yespo CDP Began

Starting from 2014, Foxtrot has been developing personalized marketing in collaboration with Yespo—a partner capable of supporting the retailer’s scale and goals. At that time, they implemented the first communication channel—email, which quickly proved its effectiveness. Later, Foxtrot scaled customer interaction by adding new channels and functionality:

Over 11 years of collaboration with Yespo, Foxtrot has:

• integrated 8 communication channels (Email, Widgets, Web push, Mobile push, App Inbox, In App, Viber, Telegram);
• set up 77+ triggers, 15 of which are custom, adapted to the retailer’s unique needs;
• implemented personalized product recommendations on the website and expanded them to direct channels, offline, and call center.

This way, the retailer provided millions of customers with quality personalized experience and doesn’t plan to stop at what’s been achieved.

“Yespo CDP is a maximally professional team where we have our personal customer success manager, impeccable analytics and reporting, including for our custom solutions. From idea to development, testing, and implementation—we have complete fulfillment,” emphasizes Oleg Nikolsky, CMO of Foxtrot.

Unified Customer View

A unified customer profile in CDP is not just a list of purchases and contact data. It’s a complete chronology of interaction with the brand, collected in one place. This is exactly the profile that helps Foxtrot personalize communication for each customer.

In a unified contact card, the system combines data about:

• interaction with campaigns: message delivery, opens, clicks, unsubscribes, spam complaints;
• behavior on the website and mobile app (session duration, views, searches, adding to cart), offline data from CRM (purchases, call center appeals, order parameters);
• demographics: age, gender, city, etc.

A unified customer view ensures the relevance and correctness of business communications with the consumer. For example, Jakob views tablets on the website—CDP knows the following data about him:

• demographic—39 years old, Munich, etc.;
• historical—two months ago he bought a phone;
• behavioral—last website visit from a promotional email; viewed X category of products.

The system understands that Jakob viewed tablets from a specific manufacturer in the mid-price segment, so it suggests he consider certain models that he might not have found independently, but which will likely interest him as they match his requests. Since the customer interacts well with the email channel, further communication will be primarily supported through this channel.

Thanks to CDP uniting all channels, a seamless experience is created for the customer at all interaction points. This approach helps Foxtrot not only increase sales but also build long-term customer loyalty.

Triggers—Automated Personal Customer Support

360-degree customer understanding allows the business to send messages at the right moment and through the right channel (email, Viber, push, App Inbox) to bring back the customer and stimulate purchase. To automate precise interaction with millions of customers, the retailer developed an extensive trigger map, where a specific scenario activates when the customer performs a particular action.

Each of the used PRO triggers plays an important role for the retailer, and the most effective of them:

By CTR—price reduction group. The biggest share of these triggers’ success depends on Foxtrot’s team, which sets up discounts, constantly updates them, and adds logic with promo codes. This combined with automation gives good CTR.

By conversion:

• Abandoned category: reminder about viewed products.
• Price reduction on cart items: notifications about promotions on cart items.
• Abandoned cart: call to complete the purchase.

As we can see, different triggers show the best results by CTR and CR. This depends on their direction, as some scenarios work to increase website traffic, others—to grow purchases.

"The best results come from a comprehensive system: at all stages of the sales funnel, use corresponding chains in different communication channels,"—Oleg Lesov, CPO Yespo.

Product Recommendations 2.0: Transformer Model + LLM

Two years ago, Yespo made a breakthrough in product recommendations by switching to a transformer model enhanced with large language models (LLM). This approach fundamentally changed personalization compared to previous algorithms and brought significant growth in key metrics.

How did recommendations work before?

Until 2019, Foxtrot used collaborative filtering, which was based on similarity between products and user behavior. For example, if a customer bought a television, the system would suggest accessories (like a soundbar) or similar models that were often bought together with this product. The algorithm analyzed historical data: if two products were often clicked or bought together, they were considered related. However, this approach had limitations: it didn’t consider the context of customer actions, and couldn’t effectively handle new products without sales history.

How does the transformer model work?

In 2019–2020, Yespo implemented a transformer model that analyzes the sequence of customer actions—from views and searches to adding to cart, purchases, both online and offline. This deep learning architecture allows the system to “understand” not only individual actions but also their context and relationships.

Since 2024, large language models (LLM) have enhanced the transformer model, adding to AI the ability to analyze product characteristics: categories, brands, prices, descriptions.

The new model allows creating recommendations that understand the essence of the product, not just its ID.

Previously, the machine only saw user behavior: if two products were often clicked together, they were considered similar. But now AI understands what the product is, its essence, and selects more accurate recommendations.

Key advantages of Transformer model + LLM:

• Sequence analysis: considers the order and duration of interactions.
• Contextual understanding: determines whether the customer is looking for promotional products or premium products.
• Deep connections: suggests compatible products (capsules for coffee machines, soundbars for TVs).
• New product handling: the system can recommend products recently added to the catalog.
• Working with minimal data: AI selects recommendations for products without purchase history through characteristic analysis.
• Relevant product replacement: selection of maximally analogous products for those out of stock.

How does this look in practice?

A customer views a gaming console, then searches for gamepads, and previously was interested in VR glasses. Collaborative filtering might suggest a popular game that was bought together with the console. The transformer model would register interest in interactive entertainment and suggest a specialized VR controller, even if such combinations weren’t previously bought. Thanks to LLM additionally analyzing product descriptions, the system can suggest novelties without sales history, solving the cold start problem.

Algorithm update results

The combination of transformer model + LLM gave even better results:

• +33%—CTR growth of recommendation blocks.
• ×2.08 traffic increase—for the “Personally for you” block.

“It should be noted that this wasn’t a cold start—recommendations at Foxtrot have been working for 8 years: the system has long been working with huge amounts of customer behavior data. At the same time, the algorithm has significantly improved user experience,” emphasizes Oleg Lesov.

Product Recommendations 3.0: Smart Product Categorization

It seemed recommendations were already working very well, but the Yespo team noticed that the retailer was losing opportunities through overly broad product categories. Previously, AI simply didn’t select products from the same category to avoid duplications in recommendations. This limited cross-sales when functionally different products within one category made sense for joint use. To solve this, Yespo implemented an LLM model that independently refines product subcategories and forms more accurate, relevant recommendations.

How does this work?

The LLM model analyzes the categorical tree (tens of thousands of products) and determines functional connections. The model refines subcategories and suggests compatible or complementary products. This automates selection, similar to how a category manager works.

For example, in the “Toothbrushes” category there are various products: brushes, care accessories, attachments, stands, travel cases, etc. In classic “Bought together” block logic, the system doesn’t recommend products from the same category to avoid irrelevant pairs (two identical brushes).

Thanks to recategorization, now logical additions to the purchase can be suggested to the customer: an attachment or case. The system will form relevant product pairs even from the same general category, improving personalization quality and cross-sales.

Results of “Frequently bought together” block (March-May 2025):

• +39% CTR increase.
• 1.5x growth in accessory sales.

About Collaboration Results

“We are very satisfied with the collaboration. This LLM model adds enormous value to our offerings. We increased LTV, revenue, reduced marketing budget usage, and working resources. We increased retention, number of repeat sales, doubled conversion on certain triggers. Yespo is our main strategic partner in retention, and we gladly meet each other halfway,”—Oleg Nikolsky.

What’s Next?

Foxtrot plans to maximize personalization of user experience. Key directions:

• Trigger map development: continuing to create detailed scenarios for all customer journey stages.
• Predictive segmentation: forecasting customer needs using AI, for example, probability of purchasing a certain category.
• Recommendation improvement: enhancing models considering seasonality and trends.
• Mobile app expansion: more personalized offers in the app where customers spend significant time.

More cases and useful insights—on our YouTube channel!

Summary

The collaboration between Foxtrot and Yespo is a clear example of how strategic partnership and advanced technologies can ensure quality marketing in retail. Over 11 years of collaboration, the retailer went from email campaigns to new-level personalization. Despite Foxtrot having all capabilities for independent development of any functionality, the retailer chooses Yespo as a ready solution. Since collaboration with CDP saves company time and resources, and all retailer ideas are developed and implemented by the platform’s experienced team.

Thanks to joint work between the retailer and Yespo, a unified view of 13 million customers is ensured, quality customer support through triggers and innovative product recommendation algorithms.

The presented successful implementation cases of transformer model + LLM, and then recategorization, demonstrate how AI technology development combined with Yespo CDP allows retailers to increase sales and customer engagement.

Yespo became for Foxtrot not just a platform, but a partner that offers custom solutions per client request, provides analytics and constant support from idea to implementation and results measurement. In the retailer’s plans, even deeper AI integration for predictive segmentation and personalization expansion in the mobile app, opening new horizons for growth.

Want to scale your marketing and achieve significant results like Foxtrot? Book a consultation with Yespo experts who will help transform your data into profit and customer loyalty.

Special Request Inline