Shelf Intelligence. Built for 150 Countries.

The challenge: presence without visibility

The client is one of the most recognizable names in the food and beverages industry, operating across more than 150 countries, with retail as one of its largest and most critical sales pipelines.

But scale creates its own blind spots.

With products spread across thousands of stores in diverse markets, the client had no reliable way to measure shelf visibility, track product placement quality, or identify exactly where sales were underperforming and why. Meanwhile, competition, both local and international, was intensifying across every market.

The brief was clear:

1. Build a system to monitor product visibility and shelf presence at the store level.
2. Use ML to distinguish high-performing placement from low-performing placement in real time.
3. Make it work in remote locations with little to no internet connectivity.

Starting with the shelf, not the spreadsheet

Before building anything, W2S mapped the client’s retail flow to understand the product flow cycle. How products moved through stores, how planograms were currently being defined, and where the gaps between intended placement and actual placement were largest.

The system was designed around that reality, not around an idealized version of it.

An ML tool that reads the shelf like a person would

A mobile application was built using edge computing and Core ML that lets field teams capture an image of any store shelf section and have it instantly converted into readable, structured data. It helped in identifying placement gaps, compliance issues, and visibility scores without manual input.

“The application captures what a human eye sees on the shelf and turns it into something the business can act on.”

Built to work where connectivity doesn’t

Many of the client’s retail touchpoints are in remote or low-connectivity areas. W2S built offline-first functionality into the app. All ML analysis runs on-device via Core ML and edge computing, with data synced to the central platform once connectivity is restored. No dropped data. No manual catch-up.

Planogram intelligence that feeds strategy

With shelf data now structured and centralized, the client could build targeted sales and marketing strategies around actual placement performance and not assumptions. Underperforming locations got specific interventions. High-performing placements became the benchmark.

Built on

• Backend: Python, Django Framework
• Frontend: React JS
• Mobile: React Native
• Local Database: Realm
• ML: Core ML, Edge Computing
• Cloud: Azure
• Project management: JIRA, GitHub
• Communication: Slack, Google Meet/Zoom

An 8-person team comprising UI/UX, data engineers, backend and frontend developers, a mobile developer, cloud architect, QA, project management, and a customer success manager delivered the full platform.

Launch and what’s next

With 6+ months in development, the platform is built for expansion — new markets, new store formats, and deeper ML models as more shelf data flows in over time.

Launch and what’s next

Your products are on the shelf. The question is whether they’re working for you. Let’s find out.