The rise of new retail has facilitated talent flow between the tech and retail industries. Tech people understand technology, while retail people understand operations. The collision of ideas between these two groups has produced many interesting things, such as unmanned convenience stores, unmanned shelves, and unmanned retail cabinets. If we categorize retail formats by their level of radicalness and technological content, the differences become obvious.
Interestingly, these four retail formats also differ in market share and growth rate: unmanned shelves have low market share and slow growth, seemingly hitting a ceiling; unmanned convenience stores are few in number but growing at a decent pace; traditional retail stores have large and stable market size but slow growth; only smart vending cabinets combine decent market share with fast growth, undoubtedly making them the most investment-worthy star product at this stage.
Speaking of smart vending machines, many people are familiar with them. Their predecessor, the mechanical traditional vending cabinet, emerged in Japan in the last century. Domestic vending machines started late but coincided with the proliferation of mobile internet, so they supported mobile payments from the beginning. In recent years, with the rise of new retail, using technology to reduce costs and increase efficiency in traditional retail has become a major trend, also promoting the prosperity of smart vending machines. New forms such as automatic orange juice machines, coffee machines, ice cream machines, and bento machines continue to emerge, with the basic design concept being to use technology to simulate vending services and reduce the risk of loss.
Interestingly, those who make smart vending cabinets also fall into two categories. One is the idealists who believe in technology; they are not satisfied with existing technologies like RFID and gravity sensing, and are still exploring visual recognition, such as automatic product recognition or facial recognition, attempting to make the vending machine form more open. The other is the realists who focus on market performance, more concerned with whether the product meets actual operational requirements, and cautiously adopt new technologies. Guangzhou Jumi Intelligent is a representative of this latter type.
Rejecting One-off Deals
Visual recognition is one of the most popular applications of artificial intelligence in recent years. Thanks to advances in deep learning algorithms, the accuracy of AI in image processing has greatly improved. Through data training, computers can recognize objects like humans, but in complex real-world environments, visual recognition still faces challenges. For example, in open vending cabinets using visual recognition, when people quickly and repeatedly take items, machine vision may fail to recognize them.
"Visual recognition is the future, but its current performance is unstable and needs time to mature and meet operational needs. When making products, we shouldn't show off technology; we also need practical operational solutions so that technology can truly improve efficiency."
With long hair and a cap with an alien logo, Zhu Zhentao, founder of Jumi Intelligent, has a rebellious style, but when he talks about products and the industry, you can feel his rigor.
Zhu Zhentao has started businesses across multiple industries multiple times, with both successes and failures. He spent ten years in the traditional FMCG industry, four years in internet media, and six years in smart device R&D, manufacturing, and supply chain management. Therefore, Zhu understands both the pain points of retail and the boundaries of technology, which determines that his approach to vending machines is closer to market demand.
Jumi Intelligent is a 2B company serving brand owners in the FMCG industry and channel operators and operators in the chain retail industry, providing them with "equipment + online agency operation services." That is, the smart vending cabinets developed by Jumi Intelligent are branded and operated offline by clients, after which Jumi continues to provide online agency operation services, such as personalized marketing and automatic analysis of restocking needs and strategies, rather than a one-off equipment sale.
"We position ourselves as an internet company, not a pure hardware manufacturer, and we also do software development. Under the internet model, everyone wants to do a light model, but retail is an industry that requires solid, meticulous effort." Zhu believes that precisely because many are unwilling to do heavy work, Jumi should do hardware. In fact, Jumi mainly focuses on quality control management and equipment integration in hardware, while component production relies on the complete supply chain system in South China. The team puts more effort into innovation at the system's underlying level.
Business Boundary Lies in Standardization
Regarding what the team does and doesn't do, Zhu has a basic judgment: standardization. If a service required by a customer can be standardized, Jumi will try it; if it cannot be standardized, the team stays away. Therefore, the team does B-end software and hardware development, online agency operations, offline after-sales, and equipment supply chain management, but has not ventured into endpoint operations.
"The offline environment where vending cabinets are placed is very complex, making site selection highly labor-intensive, which is not our strength," Zhu said frankly. He hopes Jumi can become an enabler, helping brand owners sharpen their knives, while the brand owners do the chopping.
Take software and hardware development as an example. In the industry, some manufacturers only do hardware and outsource software to third-party companies. Hardware companies are unwilling to open interfaces to software companies, and third-party software companies don't want to do online operations. This results in products where hardware and software cannot be integrated, making it difficult to provide online operation services, forcing them into one-off equipment sales. From a product perspective, such vending machines lack underlying sensing capabilities, only having mobile payment and inventory management, but the data between hardware and software is not interconnected, putting pressure on later operations.
Take after-sales maintenance services. For a vending machine with integrated hardware and software, when a malfunction occurs, the software layer can know the type of fault and determine whether it can be resolved remotely or requires on-site service, whether it needs immediate handling or can be deferred. This reduces time wasted by after-sales personnel on blind inspections.
Regarding product standardization, Jumi's main business previously was customizing retail cabinets for brand owners. With accumulated product and operational experience, Zhu believes it is the right time for Jumi to produce standardized products at scale.
In Zhu's view, customization is a necessary process before standardization, helping the team deeply understand market and operational rules. During this process, Zhu has seen many incorrectly defined products, mostly due to over-defining product requirements without considering actual operational issues. Therefore, as the team's understanding of market demand becomes clearer, Zhu decided to create standardized products to achieve scaled output. For customization business, Jumi will draw on previous experience to filter customization demands that align with market rules.
"New retail has awakened many retail enterprises, trying to capture offline traffic, and smart vending machines are a great tool. Retail giants like 7-Eleven and FamilyMart are also exploring the traffic business behind smart vending cabinets. We hope to serve more than 100 retail brand owners and operators in the future, deploy 200,000 devices in the market, and do a good job as a service provider," Zhu said.
Making "Cabinets" with the Mindset of Running a "Store"
Currently, Jumi Intelligent mainly offers two standardized vending cabinet products. The first is called Yunxiao Cabinet, which uses mobile phone scanning to purchase via H5 pages, and mobile scanning to unlock for restocking via a mini-program. Its main features are small size and high cost-effectiveness.
The second is a 24-hour smart unmanned store, featuring ultra-large storage capacity, about 5-8 times that of similar products. Behind this is Jumi's logic for making smart vending cabinets: positioning the vending cabinet as a "store."
Zhu noticed that when it comes to vending machines, many people's first impression is a beverage machine. This impression is mainly influenced by Japan. Over decades of evolution, vending machine endpoints in Japan have gradually concentrated in the hands of large beverage companies, which have natural advantages in supply chain and product pricing. The original channel operators gradually became professional service providers. Zhu believes this is the inevitable fate of traditional vending machines under the traditional model: the number of SKUs a vending machine can carry is small, but beverages are the most frequently purchased items in actual scenarios. To improve shelf and supply chain efficiency, merchants eventually concentrate product selection on a few beverage products, which is the inevitable result of the Pareto principle. As convenience stores increased in Japan, many vending machines eventually became machines that only sell beverages. Of course, there are also some snack vending machines, but their proportion is still too small.
However, this is the fate of Japanese traditional vending machines and does not apply to the domestic retail cabinet market. Zhu found in research that American vending cabinets did not fall into the beverage machine model; instead, snack and general merchandise machines are more common. Domestic demand is more diversified, with broader requirements for product forms. Generally, best-selling items in chain retail stores are also concentrated in 20-30% of categories, so why don't they remove the poorly selling items and reduce floor space? This actually reflects consumer interests; the brand attributes of the chain retail terminal far outweigh the attributes of a beverage machine that satisfies immediate needs.
To avoid falling into the "beverage machine" trap, Jumi's "smart store" has defined the smart vending cabinet as a "store" from the beginning, offering as rich a variety of products as possible. This maintains consumer convenience and benefits the vending machine's traffic attraction. The design philosophy shifts from maximizing operator efficiency to maximizing consumer satisfaction and convenience.
To this end, Jumi has increased product storage capacity and designed new shelf forms to accommodate as many product types as possible; on the algorithm side, it helps brand owners define "essential" products.
"Besides products with high purchase frequency, if a product is purchased by one person more than five times, even if not high-frequency, we define it as indispensable. For example, if a product is purchased by ten people within a certain period, we also define it as indispensable," Zhu introduced. "Also, offices may not have someone cut their hand every day, but we suggest clients consider placing items like band-aids and red flower oil for emergencies."
Currently, Jumi's standardized products are already in production, and the company recently secured tens of millions of yuan in Series A funding. It will build a large production base in Yongzhou, establish an online operation center, and export smart vending machine operational experience to operators.
-END-