Vinasoy reports a 20% reduction in out-of-stock rates using AWS AI
Vinasoy says an AWS-based generative AI shelf-monitoring system cut its out-of-stock rate by 20% in two months and raised monthly outlet coverage from about 17% to more than 70%.
Vietnamese soy milk manufacturer Vinasoy says a retail display monitoring system built with Renova Cloud on AWS reduced its out-of-stock rate by 20% within two months. These are company-reported figures; the source article does not provide an independent evaluation or the methodology used to measure the reduction.
From manual photo review to weekly monitoring
Previously, Vinasoy staff manually assessed store-visit photographs in its Distribution Management System. The company says each scoring cycle required nearly 2,000 work hours and more than 20 days, covering roughly 17% of outlets per month while each store was typically monitored once monthly.
In the new workflow, salespeople photograph shelves and upload the images to the distribution system. Vinasoy's image-recognition technology running on Amazon SageMaker identifies products, including 24 variants of Vinasoy milk across different packages and sizes. Amazon Bedrock then evaluates each display against approved merchandising standards and produces a compliance score.
Vinasoy reports that the system now covers more than 70% of outlets each month with weekly monitoring. Sales teams receive results within one to two days, compared with waits of several weeks for display-compliance data under the previous process.
What the performance figures mean
According to the company, the recognition component processes 22 images per second and is 1,300 times faster than manual inspection at that specific task. That figure describes image-recognition throughput, not the entire time required to assess a display and deliver a score. Vinasoy separately describes score delivery as up to 20 times faster than its previous process.
The source says the system can handle imperfect photographs and unusual shelf arrangements, but provides no numerical accuracy measure. Its performance under those conditions therefore cannot be independently assessed from the report.
Expansion into supply chain and packaging inspection
Vinasoy plans to extend display monitoring to all its outlets in the coming months. It is also applying AI-driven analytics to sales management, demand forecasting, and distribution optimisation. The company says preparation of a management sales report has fallen from about three days to roughly five minutes, a separate result from the image-processing metrics.
The next project with Renova Cloud is an AWS-based packaging inspection system. The practical value of these deployments will depend on maintaining accuracy at scale, the quality of incoming photographs, and whether the reported business metrics can be validated.

Source: AI News