4SELL scales its retail AI platform with Yandex Cloud

What this is about

4SELL owns an AI platform for brick-and-mortar retail and HoReCa, analyzing checkout conversations and POS transactions to uncover upselling opportunities. The Yandex Cloud infrastructure supports 24/7 service availability, processes 100% of transactions, and enables seamless scaling with zero capital expenditures.

Find a scalable infrastructure to ensure fault tolerance and high availability

4SELL is a B2B SaaS platform for brick-and-mortar retail and HoReCa. It analyzes employee-customer conversations, helps monitor sales quality, identifies missed upselling opportunities, and utilizes this data to boost staff motivation. 4SELL combines speech analytics, computer vision, sales management, and gamification; once introducing video analytics, it started tracking the beginning and end of conversations, the exact moment a customer approaches the checkout counter, suspicious transactions, and theft attempts.

As the product evolved, 4SELL experienced rising infrastructure requirements. It now needed resources to process audio, video, and POS transactions concurrently, while also ensuring the stable operation of mobile apps and analytics dashboards. For some clients, local data residency, low latency during real-time processing, and secure data transmission were critical. The company required an infrastructure capable of scaling under heavy workloads without capital expenditures for proprietary hardware.

4SELL evaluated multiple infrastructure providers, comparing them across managed service offerings, cost efficiency, support quality, performance velocity, and secure data transmission capabilities.

Finally, the company opted for Yandex Cloud. A critical factor was a data center in Kazakhstan, which enabled local data residency and reduced latency for regional users. Furthermore, Yandex Cloud provided the managed services required by 4SELL to scale its product without unnecessary capital expenditures: Yandex Managed Service for Kubernetes^® and Yandex Managed Service for PostgreSQL enabled automatic scaling under heavy workloads and a growing amount of data.

Built real-time analytics within Yandex Cloud

4SELL migrated its platform to Yandex Cloud independently, without involving third-party contractors. The migration spanned approximately three weeks and was run across three distinct stages:

1. First, the team deployed the Kubernetes^® cluster and databases.
2. Next, they migrated the data and active services.
3. Finally, they ran performance tests and routed clients to the new infrastructure.

The primary challenge during the migration involved configuring network rules and security groups. The team resolved this independently by consulting the Yandex Cloud help guides. Currently, their runs in production, servicing 112 licenses across 8 retail chains.

4SELL’s infrastructure is built on Yandex Cloud managed services. Yandex Managed Service for Kubernetes^® and Yandex Compute Cloud are used to run backend services, API servers, and audio stream processing components. POS transactions, employee data, bonuses, and metrics are stored in Yandex Managed Service for PostgreSQL, while conversations and analytics results reside in Yandex Object Storage. Audio files are utilized solely for transcription and are not retained after processing. Yandex Monium Metrics and Yandex Monium Logs are responsible for collecting metrics and logs, while Yandex Virtual Private Cloud provides an isolated network infrastructure and secure data transmission channels. The project also utilizes Yandex Container Registry for application storage, Yandex Certificate Manager for SSL certificate management, and Yandex Application Load Balancer for load balancing. Data safety is maintained through Managed Service for PostgreSQL and automated backups.

This is the basic 4SELL’s scenario. Audio data from checkout counters flows into the cloud environment, where the backend services transcribe the recordings, clean the output, remove redundant data, and detect upselling and cross-selling attempts.

The platform then correlates speech analytics results with data from POS systems, including iiko and r_keeper connected via API, flags successful upsells, and stores the analytical insights. Events are processed in near real-time, making them immediately available for staff incentive mechanics and business analytics. For instance, a salesperson can receive a bonus for a verified upsell or cross-sell right after the purchase, rather than at the end of the month with their salary.

Upsell growth, platform stability, and cost optimization

Following its migration to Yandex Cloud, 4SELL achieved a fault-tolerant infrastructure that supports 24/7 platform availability and processes 100% of transactions across all locations, even during monthly client base growth spikes from 84% to 102%. New environment deployments were fully automated, scaling was implemented with zero proprietary hardware purchases, and infrastructure outlays shifted from upfront capital expenditures to a month-by-month cloud resource payment framework. The platform’s gross margin stands at 60%, while its LTV to CAC ratio exceeds 3.95 and continues to rise.

For 4SELL clients, the business impact is driven by a surge in upsells and average order value. According to company data, the solution’s ROI can reach up to 650% alongside a 5% increase in average order value, with breakeven achieved through just 3 to 7 additional transactions per day. Within the Hani retail network, 4SELL was deployed across ten locations, analyzing 93,892 orders. As a result, the share of orders containing upsells rose from 3.1% to 9.0%, the average order value increased by 10% within 60 days, and upsell revenue grew 1.7 times. Employee script compliance over the same timeframe increased from 20% to 45%.

Within the Biopharma pharmacy chain, the implementation of 4SELL generated an additional KZT 500,000 per month in upsell revenue per pharmacist, which climbed to KZT 800,000 per month per pharmacist after a few months.