What an Unmanned Store Is
An unmanned convenience store lets someone shop with no staff on the floor. It is the usual form of grab-and-go retail. Three technical paths are in use. RFID: each item carries a tag, and a reader at the exit settles the basket. AI vision: several cameras and a model name the item and the action, so checkout needs no scan. Weight plus vision: a shelf scale is checked against the camera, which lifts the hit rate. Whichever path is chosen, IoT is what joins the door, the shelves, the payment, and the cameras.
Around 2017 money poured into these stores in China, and then the field shook out. What is left is a closer look at whether the technology is mature and whether one store makes money. Offices, parks, and hospitals are the main sites: the customers are regular, and the loss can be held down.
iResearch and others put China's unmanned retail market in 2023 — stores and smart cabinets together — at about 25 billion yuan, passing 40 billion yuan in 2025, with compound growth of about 25%. The 14th Five-Year Plan for Domestic Trade encourages unmanned retail. The guide on digital transformation of retail pushes physical stores onto digital systems. The Cybersecurity Law and the Data Security Law set the rules for the transaction data.
Who Builds It, and How the Paths Compare
The chain runs from the technology (vision, RFID, shelf scales) through the integrator (shelves, the door, payment) and the operator (a brand, or a property partner) to the connectivity layer. An RFID tag costs 0.2–0.5 yuan and recognises more than 99% of items, and someone has to stick the tag on. Vision has no tag cost and recognises about 95%–98%, and it wants computing power. Weight plus vision can pass 98%, at a hardware cost between the other two. One store typically has one access set, 4–12 cameras, several shelf sensors, and 1–2 payment terminals, and wants about 10–50 Mbps on 4G/5G.
Core Pain Points
Cost and the hit rate: Vision wants computing power and a model, so the hardware bill for one store is high. RFID spends money on tags and on the labor to apply them. A wrong name or a missed item becomes a dispute and a loss. Shrinkage: With nobody on the floor, a walk-out, an item slipped into a bag, or damage to the equipment is hard to stop in the moment. The door, the video record, and a credit check are how the loss is held down. A live watch and an alarm on unusual behavior are the practical line. The machines and the restock: Refrigeration, lights, and payment terminals sit in different corners and a fault has to be answered quickly. What to restock, and how often, should follow the sales. A remote watch and a restock list cut the cost of looking after the store.
Where IoT Is Used
The door: A scan in the app or a face opens it. The controller checks identity and credit with the cloud. Entry and exit are tied to what was bought. A long stay, or several trips in and out with no payment, raises an alarm.
Naming the item, and taking the money: The vision system sends the item and the action to the cloud, builds the order, and charges the account. The link has to carry the result and the payment instruction in time, or a slow network fails the settlement.
The room and the equipment: Temperature, humidity, refrigeration, and lighting are watched, and an abnormal reading alarms. Payment terminals and cameras report whether they are online, and a fault can be diagnosed from afar. One platform looks after all of them.
Standards and Terms
- Grab and go: shopping with no cashier.
- RFID (radio-frequency identification): a tag read by radio.
- SKU recognition: naming the product, by camera or by RFID.
- Shrinkage: stock lost to theft, damage, or expiry, as a share of inventory.
One store in a technology park uses vision. It is about 30 m², with about 300 SKUs, 80–120 orders a day, a recognition rate of 97.2%, and shrinkage held within 0.8%. The door, the cameras, the shelves, and the payment terminal share a 4G link, and the operator watches them from elsewhere. Restocking moved from a fixed every-three-days visit to a visit every one or two days when the sales say so.
The guide on digital transformation of retail asks for data security and a trace of the transaction. The regulations on cybersecurity classified protection require a classified-protection assessment for a system that handles payment or a face. Devices have to follow the personal-information security specification on biometric data.
Where the Market Is Heading
Vision, RFID, and shelf weight are being used together so the hit rate rises. The range is moving from convenience goods into fresh food and medicine. A staffed store is adding an unmanned checkout so the labor bill drops. Industry standards and a shared protocol are filling in.
Summary
An unmanned store is retail and IoT in the same room. The door, the recognition, the payment, and the cameras all depend on a link that holds, which is what the visit feels like and what the operator can run.
SimtoLink supplies a 4G/5G connection for these stores, so several devices in several kinds of site sit on one access, stay online, and can be looked after from one place.