Multi-Market AI Visibility: Deconstructing Display Scoring

13/10/2026

Tuesday Afternoon at Headquarters

The atmosphere in the 12th-floor meeting room was tense. On the left, the Trade Marketing Lead presented a thick Excel report on display performance across stores in Hanoi and Ho Chi Minh City. On the right, representatives from the Thailand and Philippines offices strongly opposed this method. They argued that the process of sending photos via email and manually entering data was too time-consuming and prone to high error rates when expanding into new markets. In the center of the table, the Operations Director looked at both sides and posed the core question: How can we build a robust AI Visibility system that standardizes display scoring without bottlenecking daily reporting workflows?

This is not purely a technology problem. It is an operational challenge for a multinational retail enterprise looking to apply AI Vision to monitor product presence. To reach a consensus, we need to dissect each stage of this process, from photo capture to business decision-making.

Stage 1: Data Collection and the Reliability Challenge

Before AI intervention, the process begins with the field sales team. Sales representatives photograph the shelves and send them in. The issue arises immediately: photos are often blurry, taken at skewed angles, or worse, are screenshots of previous phone screens. When expanding into Mexico or the Philippines, this situation becomes more severe due to differences in work habits and network infrastructure. If the AI system cannot detect fake images, the entire dataset on product coverage becomes meaningless. This is the first bottleneck to resolve: How do we ensure that every image fed into the AI Vision system is honest evidence from a specific point in time? If this step is not handled well, all subsequent analysis is merely fictitious data. Humans are still needed here to retrain staff, but technology must be the automatic filter, not a tool for humans to check each image individually.

Stage 2: Automated Recognition and Standardization of Criteria

Once valid images are secured, the next step is recognition. In the past, two people spent eight hours a day just counting SKUs and checking positions. Now, AI can process thousands of images in seconds. However, the challenge lies in standardizing display scoring criteria. Each market has a different planogram (shelf layout diagram). A product in a "passing" position in Vietnam might be a "failing" position in Thailand if visibility standards are higher. The AI Visibility system must be flexible enough to load different criteria sets for each region without requiring reprogramming from scratch. This is a significant trade-off: The initial setup cost for recognition rules is high, but it saves twice the manual work time compared to humans. More importantly, it eliminates subjectivity in scoring. Humans are prone to bias or fatigue when scoring hundreds of images, but machines are not. This requires close coordination between the technical and Trade departments to clearly define what constitutes the "correct position".

Stage 3: Exception Handling and the Role of Humans

This is the stage many businesses overlook when thinking about AI. No matter how accurate AI Vision is, it will encounter exceptions. Damaged shelves, products obscured by promotional banners, or newly launched SKUs not yet in the training data. In such cases, the system needs a mechanism to flag these instances for human intervention. In practical workflows, about one-third of initial reports require re-verification by Trade Analysts. This is not a failure of AI, but a necessary complement. Humans excel at understanding context. If AI detects a product in a misaligned position, humans will assess whether it is due to the timing of the photo or an actual error by the sales staff. Maintaining a human feedback loop for abnormal cases helps preserve system reliability. If AI is allowed to make 100% automated decisions without a review mechanism, the business loses the ability to adjust when the market changes rapidly.

Stage 4: Data Integration and Multi-Market Decision-Making

Data from the three markets—Vietnam, Thailand, and the Philippines—now resides in a single dashboard. This is the true value of AI Visibility: the ability to cross-compare. Directors can see that planogram compliance is higher in urban areas than in rural ones, but this level is not consistent across countries. In Mexico, the issue might not be display position but product availability on the shelf. The system does not just say "product is missing"; it distinguishes whether it is out of stock or not placed on the shelf. This allows departments to act correctly. Trade Marketing will focus on improving visibility, while Supply Chain will focus on ensuring supply. If these departments still use siloed data, they are prone to blaming each other. Consolidating data into a single source, processed by AI, creates a common language for the entire organization. This is the shift from "managing by intuition" to "managing by real-time data." However, this requires robust and secure data infrastructure, especially when handling competitive intelligence and trade secrets.

Stage 5: The Improvement Loop and Staff Retraining

The process does not end with the report. It begins again at the training step. When AI detects that staff at a specific store chain frequently violate display rules, the system sends an alert. But more importantly, this data needs to be used to retrain the sales team. Instead of punishment, businesses can use real examples from AI cameras to objectively point out right and wrong. This is how AI supports human development, rather than replacing them. The process needs to be reviewed quarterly to update new SKUs or changes in marketing strategy. If the AI Visibility system is not updated, it will quickly become obsolete. Humans remain at the center of this loop: they provide feedback, they understand the market, and they decide on process changes. Technology is merely a tool to amplify human capabilities. Investing in staff training to help them read and use AI data is just as important as investing in software. Many businesses fail because they buy expensive technology but their employees do not know how to fully exploit its potential.

The Final Fork: Choosing a Partner

Operating AI Visibility across multiple markets is not a one-time project, but a continuous improvement process. Businesses need a partner that not only provides software but also understands the complex operational context of Asian and LatAm markets. In Vietnam, AIVISION has partnered with major corporations like Masan, Sabeco, and P&G in deploying point-of-sale monitoring solutions. Market-specific expertise and flexible integration capabilities help shorten deployment time and minimize risks. Ultimately, the decision between success and failure does not lie in the accuracy of the algorithm, but in the organization's readiness to change work processes and accept a data-driven approach. Technology is only effective when it is embedded into the daily work culture.

AIVISION helps enterprises turn AI into working systems. Explore our enterprise AI solutions, read more on the AIVISION blog, or talk to our team about your own use case.