Shelf Scoring Costs: Fixed Cameras vs. Mobile Apps
22/09/2026
Nearly half of the retail businesses I have advised over the past year believe that installing fixed cameras is the only viable path to shelf analytics. While this figure may sound credible, it creates a significant misunderstanding of the core issue. Many assume that having an eye is enough to see, but in reality, visibility must be paired with data processing and human execution. If you focus solely on initial hardware costs, you will overlook operational and training expenses, which often carry a much heavier weight.
The market is shifting rapidly. The era of chasing the latest technology is over. Today, the question is no longer 'which technology is more advanced,' but rather 'which method best fits our scale and work culture.' To understand this, we need to separate the perspectives of the three groups directly affected: leadership, operations, and IT.
Leadership Perspective: Investment or Expense?
For leadership, shelf analytics is not just an IT project. It is a margin control tool. A minor display error, replicated across hundreds of stores, can erode net profit in a silent but persistent manner. Executives are typically most concerned with ROI (Return on Investment) and payback period. They need to understand that deploying shelf scoring is not a one-time expense, but an investment in transparency.
Many mistakenly believe that fixed cameras are low-cost because they do not require smartphones or wearable devices for staff. In reality, camera maintenance, server storage, and especially the labor costs for monitoring and handling system alerts are very high. If the system generates too many 'false positives,' leadership will lose trust in the system within months. This is the biggest risk when choosing a technology solution without considering the human element.
In neighboring markets like Thailand and the Philippines, we see a similar trend. Multinational corporations are returning to 'hybrid' solutions. They no longer attempt to impose a single global standard on local markets. They accept that efficiency comes from flexibility, not absolute uniformity. For Vietnamese businesses, this means you are empowered to choose the method that best fits your workforce, rather than following foreign models.
The Trade-off Between Control and Cost
Fixed cameras allow for 24/7 monitoring. However, they generate massive amounts of data that often go unused. In contrast, the mobile app method only records data when someone scans. Is that enough? The answer depends on the frequency with which staff pass through the display area. In high-traffic areas like convenience stores, an app can capture a sufficient sample. But in warehouses or cold storage facilities with low foot traffic, cameras may be more effective.
Leadership needs a clear answer: Which method helps me sustainably minimize inventory waste and increase cross-selling? The answer does not lie in the type of device, but in the process of extracting data from that device. An expensive camera system whose data no one reads is less valuable than a free app whose data is analyzed and acted upon promptly.
Operational Reality: Who Bears the Burden?
This is the part often overlooked in financial reports. The operations team is the one directly handling the goods; they bear the most pressure when a scoring system is introduced. If the system is too rigid and inflexible, they will find ways to bypass it. When staff bypass the system, the data you collect becomes distorted. That is when shelf analytics becomes a burden rather than a support tool.
Wearable devices seem convenient, but in the Vietnamese retail environment, they face significant cultural barriers. Many employees feel excessively 'monitored,' as if wearing an electronic bracelet. This leads to higher turnover rates or passive resistance among younger staff. In contrast, mobile apps leverage devices employees already own. They reduce the feeling of being controlled because staff have autonomy in taking photos and adding notes. However, mobile apps rely heavily on image quality. Lighting in the warehouse, camera angles, or staff laziness directly affect the accuracy of computer vision.
Fixed cameras are not affected by staff laziness. They record everything. But the issue lies in 'who does what' after the camera records. Without a clear process, camera images will sit idle on the server. The operations team has no tools to view, respond, or improve. They only know they are being 'scored' without knowing exactly where they went wrong. This is the fatal weakness of a pure camera solution if it lacks a good user interface (UI) for employees.
In projects we have implemented with lubricant and instant noodle companies, I have observed a pattern: The more a method reduces friction in employees' daily work, the higher its success rate. Employees do not need to 'live and die' by data entry. They just need to work, and the system automates the rest. This is why the combination of cameras (for overall monitoring) and apps (for staff interaction and confirmation) is becoming a more popular choice than picking one or the other.
IT Team: The Hidden Pain Behind Shiny Numbers
For the IT team, the question is not 'does the image look good,' but 'is the system robust.' Each method brings its own technical pain points. Fixed cameras require stable network infrastructure, high bandwidth to transmit video to the server, and massive storage space. If the in-store WiFi is unstable, the entire monitoring system goes blind. This is an operational risk that many IT managers have not fully anticipated.
Wearable devices pose challenges regarding durability and security. Devices are prone to damage from impact, sweat, or drops. Managing hundreds of wearables for staff nationwide is a logistical and IT nightmare. Who will change the batteries? Who will update the software? If a device is lost, how do you ensure data does not leak? These questions cause many CIOs to reject the wearable approach, despite its modern appearance.
Mobile apps are technically much 'lighter.' Data is pushed to the cloud when a connection is available. There is no need to invest in a dedicated server for each branch. However, the challenge lies in integration. A shelf scoring app needs to connect with ERP, WMS (Warehouse Management System), and CRM. If these systems do not speak the same language, the IT team will have to write thousands of lines of middleware code just to connect the data. This is where 'sunk' costs hide. They do not appear in the hardware quote, but they consume months of developer effort.
Furthermore, input data quality directly affects AI model quality. If images from the app are blurry or missing angles, the computer vision model's accuracy will degrade over time if not retrained. The IT team must balance letting the model 'reject' poor images with not bothering staff too much. This balance is difficult and requires a technology partner with real-world experience, not just a software vendor.
AIVISION has partnered with major corporations like Masan and TTN to build these complex integration layers. We have learned that custom AI software is not about rewriting from scratch, but intelligently connecting existing pieces to create a seamless data flow. That is where real value is created.
The Intersection: When Technology Serves People
Ultimately, all debates about cost and efficiency become meaningless if we forget the ultimate goal: People. Shelf analytics is not meant to punish employees. It is meant to help them work more effectively. A good system must be smart enough to understand context. For example, if a product is out of stock, the system should not just flag a 'display error,' but report 'restock needed.' This difference is small, but it completely changes how the operations team receives the information.
In the Vietnamese context, where workforce turnover is rapid, system flexibility is key. You cannot build a rigid system for tomorrow if your employees change companies after three months. The mobile app method has a significant advantage here because it is easy to transition, easy to train, and less dependent on physical infrastructure. However, it requires strong backend AI support to automate the scoring process, rather than letting staff self-score (which often leads to bias or errors).
We often advise businesses not to try to choose a single method for the entire chain. Start small. Choose one area, one key product, and test all three methods. Compare monthly operational costs, not just hardware purchase prices. Listen to employee feedback. Then, scale up. This approach is slower, but far more sustainable than rushing to install cameras everywhere and realizing the waste a year later.
The Verdict on the Initial Figure
Returning to the 'nearly half' figure at the beginning. Were those businesses wrong to choose cameras? Not necessarily. They were wrong because they chose cameras as a standalone solution, rather than part of an ecosystem. If they combined cameras for overall monitoring and apps for detailed interaction, the results would differ. If they invested in the data analysis layer and staff training, that cost would be recouped through reduced waste.
The market does not reward those who chase the latest technology. It rewards those who clearly understand what they are doing. Shelf scoring is not a race for camera resolution or complex AI algorithms. It is a race for the ability to convert data into action. And that action, ultimately, must still come from people.
Every company hits this differently, and the hard part is usually the data rather than the model. To pressure-test your case quickly, talk to AIVISION - or first see how we deploy and what we have written before.