Case Study
Coke & Go
Retail / AI-Powered Smart Vending
AI-powered smart coolers using computer vision for grab-and-go beverage purchases, deployed by CocaCola Europacific Partners across New Zealand, US, Ireland, Singapore, and Europe.
Coke & Go
Tech Partnership
Retail
6 Months
- AI-powered smart cooler platform
- Computer vision product recognition and cashless payment.
- Faster checkout experience
Key Outcomes
Delivered AI-powered smart coolers with computer vision product detection, enabling grab-andgo purchases that complete in under 10 seconds from QR scan to door close
Deployed across multiple CCEP markets including New Zealand, United States, Ireland, Singapore, and Europe, with real-time inventory tracking and automated restocking alerts
Created a cashless, contactless vending experience with loyalty rewards via the Vend swift app, including a buy-10-get-1-free program driving repeat engagement
The Challenge
Before partnering with Iottive, Coke & Go was facing several issues:
Traditional Vending Machine Friction
Conventional vending machines require button selection, cash or card insertion, and mechanical dispensing-a 30–60 second process that deters impulse purchases in high-traffic locations like transit hubs, universities, and offices.
Inventory Visibility Gap
Traditional vending machines report stockouts only when a consumer encounters an empty slot.
Operators restock on fixed schedules rather than actual demand, leading to both lost sales (empty popular items) and waste (expired slow-moving items).
Product Identification at Scale
A smart cooler containing 50+ beverage SKUs — cans, bottles, multipacks of varying sizes and designs — requires reliable computer vision that correctly identifies what the consumer took, even when products are partially occluded, rotated, or from new product lines not in the original training set.
Product Identification at Scale
A smart cooler containing 50+ beverage SKUs — cans, bottles, multipacks of varying sizes and designs — requires reliable computer vision that correctly identifies what the consumer took, even when products are partially occluded, rotated, or from new product lines not in the original training set.
Our Solution
Iottive delivered a complete AIoT solution under a full-cycle product development model.
Bluetooth/IoT Expertise
Integrated IoT sensor arrays within cooler units — door sensors, temperature monitors, weight sensors, and cameras — connected via edge computing hardware that processes transactions locally and syncs to the cloud for inventory management and analytics.
Design Thinking Approach
Mapped the consumer journey from approach to departure, reducing the entire purchase to three steps: scan QR / tap card → open door → take product and close door. No product selection, no payment confirmation screen, no receipt printing — the system handles everything automatically.
Project Planning & Execution
Market-by-market rollout starting with New Zealand as the pilot, expanding to subsequent CCEP markets. Each deployment phase included local SKU training for the computer vision model, payment integration with regional processors, and compliance certification.
Functionalities Delivered
QR code and contactless card authentication, computer vision product detection (camera + weight verification), automatic payment processing on door close, Vend swift loyalty app with buy-10-get-1 rewards, real-time inventory dashboard for operators, smart restocking alerts based on actual consumption patterns, and temperature monitoring for food safety compliance.
Implementation Highlights
Sensors & Hardware
Multi-camera array inside the cooler for product identification, shelf weight sensors for verification, door sensors for session management, and temperature sensors for food safety. Edge computing unit processes vision inference locally.
Edge & Connectivity
On-device AI inference eliminates round-trip latency to cloud servers, completing product identification in <500ms. Cellular or Wi-Fi connectivity syncs transaction data, inventory levels, and software updates.
Firmware & Performance Optimization
Edge AI model optimized for the cooler’s embedded hardware, running inference at 15 FPS with 99%+ product identification accuracy. Model retraining pipeline enables new SKU onboarding within 48 hours of product launch.
Data Analytics & Visualization
Operator dashboard showing real-time inventory per cooler, consumption velocity per SKU, peak purchase times, and route optimization for restocking drivers. Demand forecasting models predict stockouts 24–48 hours in advance.
Dashboard & UX
Consumer-facing Vendswift app with loyalty tracking, purchase history, and nearby cooler finder. Operator portal with fleet-wide cooler management, alert configuration, & performance analytics.
Security & Compliance
PCI-DSS compliant payment processing. GDPR and CCPA compliant data handling across all markets. Camera data processed locally and purged after transaction — no consumer images stored or transmitted to the cloud.
Results & Impact
- Purchase completion time reduced from 30–60 seconds (traditional vending) to under 10 seconds
- 99%+ product identification accuracy across 50+ SKU variations per cooler
- Smart restocking alerts reduced out-of-stock incidents by an estimated 35% compared to fixed schedule restocking
- Vendswift loyalty program’s buy-10-get-1 offer drove measurable repeat purchase behavior across all markets
- Deployed across five CCEP markets with a scalable architecture supporting continued international expansion
“The Coke & Go coolers handle everything — the customer just opens the door, grabs their drink, and walks away. Iottive’s computer vision team delivered 99%+ accuracy across our full SKU range, and the edge processing means it works even when the cooler’s internet connection
drops. The restocking analytics alone have changed how our route drivers plan their days.”
— Karan Bakshi, Coca-Cola Europacific Partners
Lessons & Best Practices
Process at the Edge
Cloud-dependent product identification adds latency and fails during connectivity drops. On-device inference ensures every transaction
completes regardless of network conditions.
Design for the Consumer’s Impatience
Every additional step in the purchase flow reduces conversion. Eliminating product selection and payment confirmation — letting the system handle both automatically — doubled throughput at high traffic locations.
Plan for SKU Volatility
Coca-Cola launches and rotates products frequently. A retraining pipeline that onboards new SKUs within 48 hours prevents the vision model from becoming stale.
Build Local Compliance from Day One
Attempting to retrofit GDPR compliance after US deployment would have delayed the Europe launch by months. Building privacy-by-design (local processing, no image storage) from the start enabled parallel market launches.
Technology
Comprehensive IoT Technology Stack
From devices and connectivity to cloud, apps, and security — we leverage a full-stack IoT ecosystem to build scalable, secure, and future-ready solutions.
Cloud analytics
ios
Android
Why Iottive’s the Right Partner
- Computer vision + IoT integration: Iottive combines edge AI inference with IoT sensor arrays, solving the specific challenge of accurate product identification in constrained retail environments
- Multi-market deployment: experience launching connected retail solutions across diverse regulatory environments with market specific payment integrations
- Retail analytics expertise: building operator dashboards that translate IoT sensor data into actionable business intelligence for restocking, demand forecasting, and route optimization
Next Steps for Coke & Go
Expand to 10 additional CCEP markets with automated SKU catalog management per region
Add personalized product recommendations
displayed on the cooler’s door screen based on
purchase history
Integrate dynamic pricing capabilities for
time-based promotions (e.g., happy hour discounts after 5 PM)
Develop a sustainability dashboard tracking
reduced waste from optimized restocking and
energy-efficient cooler operations