Industry

FMCG

Technology

AWS Bedrock
AWS SageMaker AI

From 20,000 to 100,000+ Stores: Rethinking Retail Execution with AI

In FMCG, having products on the shelf is only the first step. Brands also need to ensure products are visible, correctly positioned, and consistently displayed across stores.

For Vinasoy, monitoring display compliance across a large retail network was a highly manual process. A dedicated team reviewed store images against display standards, limiting coverage to around 20,000 POS per month while feedback could take 20+ days.

To scale this process, Vinasoy partnered with Renova Cloud Vietnam to build an AI-powered retail display scoring solution on AWS. Combining Computer Vision on Amazon SageMaker AI with Generative AI on Amazon Bedrock, the solution automates product recognition, image analysis, compliance evaluation, and scoring.

The impact goes beyond faster image processing. Vinasoy can now monitor 100,000+ POS per month, increase monitoring frequency from around once to 4-5 times per month, and receive evaluation results in approximately 1-2 days.

This helps Vinasoy move from periodic manual audits to a more scalable, data-driven approach to retail execution, giving sales teams faster feedback and better visibility into in-store performance.

“At Vinasoy, digital transformation doesn’t stop at applying new technologies, but is a continuous journey to improve operational efficiency and create more value for customers. Implementing AI in display photo scoring is a concrete step to help the company enhance data-driven management capabilities, improve operational transparency, and optimize sales performance. More importantly, all of Vinasoy’s digital transformation efforts are aimed at better serving our customers.”

— Mr. Le Ba Be, National Sales Director of Vinasoy

>>>Let’s see Vinasoy and Renova Cloud’s collaboration

The Challenge: When Manual Retail Auditing Can’t Keep Up

Vinasoy has established internal standards for how its products should appear in stores, including:

  • Display positioning: Products should be placed where consumers can easily see and access them.
  • Product layout: SKUs should follow a consistent arrangement across displays.
  • Display quantity: Shelves should maintain an appropriate level of product presence for the available store space.

These standards are important for maintaining consistent brand visibility and execution across a large retail network.

But as the network grew, manual auditing created 3 practical challenges.

1 Limited monitoring coverage The manual process could review approximately 20,000 POS per month, leaving a significant portion of the retail network outside regular monitoring.
2 Slow feedback cycles Display images could take 20+ days to be reviewed and evaluated.

That meant teams were often looking at what had happened weeks earlier rather than what was happening in stores now.
3 Difficult to scale consistently Manual assessment also introduced the possibility of differences in how individual reviewers interpreted display requirements.

For a nationwide network, applying the same standards consistently is critical – not only for management visibility, but also for creating a fair and transparent evaluation framework for stores and sales teams.

The Solution: AI That Understands What’s Happening in the Store

Vinasoy and Renova Cloud Vietnam implemented an AI-powered display image scoring system on AWS, combining computer vision, machine learning, Generative AI, and scalable cloud infrastructure.

The solution automates the journey from store image → product recognition → compliance analysis → scoring.

AI-powered image understanding

The system uses custom computer vision models hosted on Amazon SageMaker AI for tasks including:

  • Object detection
  • Image segmentation
  • Image embedding
  • Vinasoy SKU recognition

Amazon Bedrock is used for Generative AI capabilities, including contextual evaluation, multimodal reasoning, and more complex scoring logic.

>>>Explore AI solutions with Amazon Bedrock 

This architecture allows different AI capabilities to work together rather than treating image recognition as a standalone task.

Solution architecture design [Source: Renova Cloud]

Image 1. Solution architecture design [Source: Renova Cloud]

From images to actionable compliance scoring

The operating workflow is straightforward for the field team:

Visit the store Sales representatives visit the store and check inventory.
Arrange the products Products are displayed according to Vinasoy’s merchandising standards.
Capture images The representative photographs the completed display.
AI processes the images The images move through Vinasoy’s data environment and AWS-based processing architecture.
Automated evaluation The system recognizes products, evaluates the display against predefined criteria, and generates a score.
Management receives the result Vinasoy can use the evaluation to monitor execution and identify areas requiring improvement.

Staff is taking pictures of products on the shelf to check on the platform. [Source: Vinasoy]

Image 2: Staff is taking pictures of products on the shelf to check on the platform. [Source: Vinasoy]

Designed for parallel processing at scale

To process more store images efficiently, the system can check multiple POS displays at the same time instead of one by one.

During testing, it handled 10 POS checks at once, taking about 4 seconds per unpaid display. For production, it can scale to 20-40 checks at once, with a target of about 1-2 seconds per POS, depending on data transfer and overall system capacity.

Important Distinction: What the 1,300× Speed Improvement Means

The widely reported 1,300× speed improvement should be interpreted precisely.

It refers to the SKU detection model inference speed: approximately 22 images per second, compared with around one minute for a person to manually evaluate an image. It does not represent the end-to-end scoring workflow, nor is it solely attributable to parallel processing.

The complete workflow includes compliance validation, planogram matching, and GenAI-based scoring, which require additional processing.

This distinction is important because the real transformation is not simply “AI processes images faster.” It is the ability to build a scalable, automated evaluation pipeline around those AI capabilities.

According to the Deputy CEO of Renova Cloud

“Technology only creates true value when it drives transformative change in business operations. At Vinasoy, we see a powerful combination of innovative thinking and strong execution capability in their operational digitalization strategy. The AI solution deployed on the AWS platform enables the company to make better use of existing data, accelerate information processing, build a consistent evaluation framework, and progressively move toward a real-time, data-driven operating model.”

Vinasoy and Renova Cloud Vietnam have launched an AI-powered display image scoring system on the AWS platform.

Image 3: Vinasoy and Renova Cloud Vietnam have launched an AI-powered display image scoring system on the AWS platform.

Business Outcomes

AI has helped Vinasoy move from manual, periodic display auditing to a faster, more scalable and data-driven approach to retail execution – giving sales teams more timely visibility and helping them maintain better execution across the retail network.

What changed Before AI With AI Business impact
Monthly monitoring coverage ~20,000 POS/month 100,000+ POS/month 5× more stores monitored, giving Vinasoy visibility across a much larger share of its retail network
Monitoring frequency ~1× per month 4–5× per month Moves from periodic audits to weekly visibility, allowing teams to identify and address display issues sooner
Evaluation turnaround 20+ days, sometimes nearly 1 month ~1–2 days Faster feedback makes display data more useful for day-to-day sales execution
Compliant display execution ~1–2× per customer/month ~4–8× per customer/month Encourages more consistent display execution across store visits
SKU recognition Manual image assessment ~22 images/second Automates a time-consuming part of image review and enables the process to scale with retail volume

>>>Explore more how AI solutions can apply in any use case

Built to scale with Vinasoy’s growing network

The solution is designed to support a significant increase in image volume as deployment expands.

Current scale Target scale
8,000 key-shop POS → ~24,000 images

51,000 non-key-shop POS → ~153,000 images

59,000 POS in total → ~177,000 images

 20,000 key-shop POS → ~60,000 images

100,000 non-key-shop POS → ~300,000 images

120,000 POS in total → ~360,000 images
The estimates are based on approximately three images per POS.

Future Plans

Technology is applied to ultimately serve the goal of improving operational efficiency and delivering more value to customers. [Source: Vinasoy]

Image 4: Technology is applied to ultimately serve the goal of improving operational efficiency and delivering more value to customers. [Source: Vinasoy]

Vinasoy is expanding AI beyond retail display compliance across distribution, marketing, sales, customer care, and quality control.

  • AI-powered customer care: Reduced average response time from 17 minutes to 21 seconds and enables 24/7 support.
  • AI Host for livestream: Supports livestream sessions when human hosts are unavailable, achieving approximately 1.5× investment effectiveness.
  • Agentic AI: Vinasoy and Renova Cloud plan to explore AI that can analyze data, recommend actions, and support business workflows across distribution, marketing, and customer service.

The vision is to move from:

See → Understand → Score → Recommend → Act

The display-scoring solution provides the foundation by turning store images into structured business data that can support faster decisions and more intelligent operations.

Key Takeaway

Vinasoy’s AI journey is not just about processing images faster. It is about turning a manual process into a scalable, data-driven operating capability.

With Computer Vision, Generative AI, and cloud technology, Vinasoy can monitor more stores, more frequently, and respond faster to in-store execution issues.

From “We check what happened” to “We see what is happening – and act sooner.”

Ready to Transform Your Business with AI?

From retail execution to customer experience and beyond, AI can turn complex, manual processes into scalable, data-driven operations. Work with Renova Cloud to identify the right AI opportunities, build practical solutions, and turn them into measurable business impact.

Ready to explore what’s possible? >>>Work with us and >>>Find more AI solutions today.