XenonStack collaborated with a global premium spirits manufacturer to revolutionise retail outlet analytics through a cloud-native, AI-driven platform built on AWS. Manual inefficiencies, inconsistent data, and delayed insights plagued the legacy system. XenonStack’s solution automated image capture and analysis using advanced computer vision models, integrated real-time dashboards, and deployed a fully managed architecture compliant with global data regulations.
The result was a scalable, secure, and intelligent platform that significantly reduced the need for on-site visits, improved data consistency, and delivered real-time actionable insights. By leveraging Amazon EKS, YOLOv8, and a range of AWS services, the platform elevated retail execution strategy, reduced labor overhead, and enhanced cross-market performance visibility. This case study reflects XenonStack’s leadership in delivering end-to-end AWS Managed Services, showcasing deep technical acumen and a commitment to measurable business impact.
Figure 1: Retail Analytics Automation Customer Challenge
The client, a globally recognised premium beverage producer, faced serious inefficiencies in retail outlet execution analysis. Traditional approaches involved periodic manual audits, often requiring physical site visits across dispersed geographies. These costly, inconsistent methods often resulted in delayed performance reporting, impacting market responsiveness.
Operational bottlenecks stemmed from a lack of a unified digital system for collecting and interpreting retail shelf data. The manual process was labour-intensive and error-prone, leading to governance issues, unreliable KPIS, and fragmented insights. The client needed a comprehensive platform to automate data collection, ensure high data fidelity, and empower stakeholders to make timely, informed decisions across global retail markets.
Customer Information
-
Industry: Beverage / Spirits Manufacturing
-
Geography: Global operations with retail presence in North America, Europe, and Asia-Pacific
-
Company Size: Large Enterprise
-
Key Objectives: Automate outlet analysis, reduce operational costs, improve reporting accuracy, and enhance strategic retail execution
Business Challenges
The client’s legacy operations relied on decentralised, spreadsheet-based tools and manual inputs to track in-store performance and brand compliance. This approach led to:
-
High labor costs due to frequent on-site visits for shelf audits
-
Inconsistent data due to varied regional practices and manual reporting errors
-
Delayed insights, making real-time market responsiveness unattainable
-
Stock-out incidents that impacted brand presence and sales opportunities
-
Siloed data structures that hampered holistic, cross-functional analysis
-
Lack of a feedback mechanism for field teams, causing misalignment in brand execution
The business needed an intelligent, cloud-native analytics platform capable of automating visual data capture and analysis to transform its retail operations into a more proactive, insight-driven process.
Technical Challenges
The technical solution needed to address several complex requirements:
-
High-Volume Image Processing: Thousands of shelf images require daily analysis using computer vision models to detect brand presence, pricing, and compliance indicators.
-
Accuracy and Model Performance: Achieving high accuracy (>95%) in KPI extraction demanded extensive model training and continuous optimisation of YOLOv8-based algorithms.
-
Geospatial Validation: Field data required real-time location validation to confirm the authenticity of the images submitted from retail outlets.
-
Security & Compliance: Global operations mandated GDPR compliance, data encryption at rest and in transit, IAM-based access controls, and auditability.
-
Scalability & Availability: The solution needed to maintain high performance during peak loads, using Kubernetes-based orchestration across multiple availability zones.
-
Real-Time Visualisation: Building dual dashboards for executive and admin users to track live KPIS, alerts, and outlet compliance trends.
Partner Solution
Solution Overview
XenonStack delivered a cloud-native computer vision-powered retail analytics platform hosted on AWS, engineered to automate outlet audits and enhance data-driven decisions. Key capabilities included:
-
Containerised Vision Workloads: YOLOv8 models encapsulated in Docker containers to classify, detect, and extract brand-level KPIS with over 95% accuracy.
-
Kubernetes Orchestration via Amazon EKS: EKS managed clusters with autoscaling node groups ensured optimal performance and uptime.
-
Custom Task Management Backend: A secure backend enabled task assignments, verification, and data lineage tracking.
-
Real-Time Dashboards: Amazon CloudWatch dashboards gave tailored insights to different user personas—field executives, data analysts, and operations leads.
-
Secure, Compliant Architecture: Multi-layered security using IAM, KMS, and CloudTrail to meet GDPR and internal governance standards.
Architectural Diagram
AWS Services Used
-
Amazon VPC & Subnets: Isolated network segments for public and private workloads
-
Amazon Route 53: Low-latency DNS resolution
-
Amazon EKS: Orchestration of Docker-based AI models
-
Amazon ECR: Container image repository integrated with CI/CD
-
Amazon S3: Image and artefact storage with lifecycle policies
-
Amazon RDS: Relational metadata store for task management
-
AWS CloudWatch: Logging, monitoring, and dashboarding
-
AWS CloudTrail: API-level audit trails
-
AWS KMS: Encryption of data at rest
-
AWS IAM: Role-based access and policy enforcement
-
AWS Secrets Manager: Secure management of API tokens and DB credentials
-
Elastic Load Balancer: Traffic management across multi-AZ services
Results and Benefits
Business Outcomes
-
50%+ Reduction in Field Visit Costs: Automation replaced manual audits, cutting overhead significantly
-
Improved Data Accuracy: Standardized image analysis led to over 95% accuracy in KPI reporting
-
Accelerated Time-to-Insight: Market response time improved from days to hours through real-time dashboards
-
Decrease in Stock-Outs: Enhanced shelf visibility reduced out-of-stock incidents, boosting revenue potential
-
Cross-Market Transparency: Unified dashboards enabled global consistency and compliance benchmarking
-
Enhanced Employee Productivity: Sales teams focused on strategic engagement rather than data collection
Technical Benefits
-
High Scalability: EKS autoscaling supports dynamic workload demands
-
Robust Compliance Posture: KMS, IAM, and CloudTrail ensured adherence to GDPR and internal data governance policies
-
Real-Time Operations: Dashboards powered by CloudWatch significantly reduced report generation and lag
-
Optimised Resource Use: GitLab CI pipelines and S3 lifecycle policies minimise infrastructure and storage costs
Lessons Learned
Challenges Overcome
-
Model Accuracy Optimisation: YOLOv8 models required multiple training iterations and fine-tuning with diverse datasets (100–150 images per brand) to reach the target accuracy
-
Compliance and Policy Automation: Security configurations were codified into infrastructure scripts to enforce IAM, KMS, and audit logging at scale
-
Dashboard Performance: Initial slow query response was improved by implementing caching and optimising database indices in RDS
-
Middleware Compatibility: Standardised data exchange formats and containerised middleware helped integrate computer vision components and backend APIS seamlessly.
More Ways to Explore Us
AWS Panorama for Edge-based Computer Vision
Building Scalable Computer Vision Solutions with Amazon
AWS Computer Vision Services and Solutions Company
