The 2018 4th 'FMCG + Internet Conference' hosted by New Distribution was held on March 18-19 at the Fairmont Chengdu Palm Springs Hotel Conference Center. It attracted over a thousand industry distributors, manufacturers, and internet companies from all over the country, with a full house and unprecedented scale. The following is the speech by Mr. Chen Haibo, CEO of DeepBlue Technology, at the conference, organized by New Distribution for readers.
DeepBlue is a team that returned to China in 2014. Overseas, we worked on dynamic recognition in machine vision. When we returned in 2014, we had only two PhDs and one postdoc. Now our postdoc and PhD team exceeds 30 people, and we are involved in autonomous driving, intelligent robots, retail upgrade, healthcare, finance, military, and transportation. Because China's new retail was hot last year, we also heavily entered new retail and launched many products. In this process, we have had many thoughts. We are B2B, empowering the industry. Perhaps many of you here will be our customers in the future. We already have Chinese and overseas enterprises as clients. Our products are exported to Europe, Japan, and Australia, and soon to the US, South Africa, and Hong Kong. We have made many products for many Chinese enterprises. In B2B, we do our best to help you succeed.
Chen Haibo, CEO of DeepBlue Technology
First, there were three retail hotspots to watch last year: convenience stores. The existence of convenience stores is mainly in the upgrade of existing stock and incremental growth. In China, the upgrade of existing stock has a large market because true first- and second-tier convenience stores have not yet entered second- and third-tier cities, and this consolidation is bound to begin. Second is incremental growth. Many second- and third-tier cities do not yet have very formal, large-scale convenience stores; they still have mom-and-pop shops, so the incremental market is relatively large. How to achieve incremental growth? Currently, opening a store doesn't make money; you work hard just to pay rent. People want to open stores but can't afford to, and if they do, they don't make money. What can be done to open stores? DeepBlue has some solutions.
The second is vending machines. We think the market for vending machines in China is huge. Traditional official explanations say the US has 6 million, Japan has 5 million, and China has 10 million. DeepBlue did due diligence on vending machines: traditional vending machines number 300,000, and current vending machines number 10 million. We think that's not enough. Pay first, then get the item; if it falls, you can't return it. We think this is not the original scenario of retail.
The original scenario of retail is that customers first get the item, they need to select, touch, smell, have contact, and then pay after holding it. If I have to pay first and can't return it if it falls, I feel a psychological barrier. If there's a store 50 meters ahead, I would walk to the store to buy things rather than buy from a vending machine. If the products in vending machines are not limited by shape, like a counter where you can sell anything, and if customers can take the product first and pay later, we think the future stock of vending machines in China will be 30 million, not 10 million.
How to achieve this? DeepBlue has launched many products, and orders are very good. Our machine vision vending machines are basically priced below 5,000.
Office shelves were also an important scenario last year. We think this scenario is correct because it brings products infinitely close to the customer group. Why did blockchain do so well? Because it is decentralized. Think about how many people go to city centers? How many young people go to Wanda Plaza? No. We recently saw data that a few years ago, a person went to hypermarkets like Carrefour several times a month; now it's halved. But they might go to convenience stores three times a day. What does that show? Infinite proximity to customers. E-commerce brings products infinitely close to customers; office shelves bring shopping scenarios infinitely close to customers. That is correct. Of course, there are problems, such as limited product variety. Can it be solved? DeepBlue can. So we also consider this a hotspot in retail last year.
Another hot concept last year was unmanned stores. Actually, regarding unmanned stores, DeepBlue has our own view. Is the unmanned store right or wrong? From our initial due diligence to many later scenarios, as an industry enabler, we have summarized some technological applications. There are three schools in the market. The first is the more traditional internet school: nothing at all, but customers scan codes to check out themselves, testing their conscience, assuming every customer is a saint. The second school is gravity sensing. Gravity sensing recently cannot adapt to all scenarios. For example, if everything weighs 500 grams, there's a problem of which one you took. DeepBlue has always said AI changes the world, AI changes retail. AI has three major parts: first, the construction of the brain; second, the eyes, where cameras replace eyes to make judgments; third, natural speech recognition. AI must have AI technology. Machine vision can solve this problem, so DeepBlue has been working hard in this field.
What is DeepBlue's idea in smart retail?
First, we think smart retail needs a brain, like AlphaGo, that can think. Although each of our cabinets has the intelligence of a five- or six-year-old, it can learn slowly. You can take things from it the first time, but after continuous iteration, you can't take things away. Machine vision is also continuously learning.
Who needs unmanned stores?
This question seems hard to answer. Does the C-end need unmanned stores? Or does the B-end need unmanned stores? I think the C-end doesn't need unmanned stores. If I am a customer, I take money and walk into a store with no service; I don't seem willing. I think the B-end seems to like unmanned stores, but that also seems wrong. So suppose next to these unmanned stores is the most low-end mom-and-pop shop. If you stand between the two stores, which one would you walk into? I think many people might say I'd still go to the mom-and-pop shop. You see, that's a problem.
The second question is that customers want to complete the simplest, fastest, and quickest process when buying. But unmanned stores often have more steps than staffed stores. When I go to a staffed store, I take an item, give money, get change, or scan a code with my phone—about three to five steps. In an unmanned store, it might take more than ten steps. Is it simpler or more convenient? We think unmanned stores seem to have some problems.
In the retail process, we need to think about what customers need and what their demands are when buying. Of course, cheap products are always important, but that's not something we can do. Customers want to complete purchases most quickly. In this process, we say checkout does not create value for the C-end in the entire retail process and should not exist; it wastes C-end time. Because it's simple: I am a C-end customer. You give me choices: first, queue to check out; second, self-checkout; third, grab and go. I would definitely choose grab and go. I tell you this store can deduct money from WeChat, Alipay, or credit card. If you find that you can't deduct conveniently because of what I took, that's your technical problem. If I don't pay, that's a legal problem. Can it be simpler? I walk into an open-front cooler and grab and go. We think it can be done for the C-end, so we have tested this.
So we say self-checkout is not new retail. Twenty years ago in the West, out of ten checkout lanes, eight were self-checkout.
Third, office shelves have recently had many problems. The model is correct because it brings products infinitely close to customers, even more advanced than e-commerce, because it brings shopping scenarios infinitely close to customers, making it more convenient for customers to purchase. But it has many problems. What is the problem? The problem is that shrinkage cannot be controlled. DeepBlue has launched some products. Each of these costs only a few thousand yuan. Many of our cooperative customers have launched thousand-yuan-level cabinets to place in their companies. Scan code or scan hand to open the door, take items, and it automatically deducts. No shrinkage. This is what we are currently promoting. Orders are OK, and our production capacity is large; we can produce about 5,000 units per day. This cabinet uses face recognition technology. Registration takes only one step. You open the door, take items, and there are cameras on the two top corners. Products can be placed arbitrarily until full. As long as you take something and close the door, it deducts. It's that simple. This is our Xiaolan series. Recent orders are large, and payment to delivery takes 35-55 working days.
How to open a convenience store without renting a space or hiring salespeople?
As mentioned, convenience stores have a huge market, but opening a store is too tiring. Renting a space, negotiating a lease takes one to two months, and you need to pay two months' deposit and three months' rent in advance, which is a lot of money. So how to open a convenience store without renting a space? This is a store of one of our recent customers. This store will appear in Shanghai and Japan. He used three of our self-service cabinets to achieve all the products of an open-front cooler and beverage cabinet in a convenience store. Each of our cabinets is about 1.35 meters, three are 4 meters. You can place products arbitrarily; there is no limit on variety. Customers find it more convenient to buy; they open the door, grab, and go. At that time, the head of Lawson Japan came and told us this is called super convenience. Why is it called super convenience? He said you have four supers: First, super convenient for customers to buy. In a staffed store, buying takes four steps; yours only takes two steps—scan code, open door, grab and go. Customers like to use it. Second, super convenient for placement. No need to rent a space; just find a wall, a wall of four to five meters, and put a few cabinets together. Third, super convenient for management. No salespeople at the front, and at the back, you save the people managing salespeople. Share one supply chain and fill it up. Fourth, super high frequency and efficiency. It uses only one-tenth of the cost of a convenience store but achieves 70% of its sales. It will also enter several market pilots every day. If you don't have a place, you can use one cabinet.
Next is the open-front cooler. This is also fun. When we walk into it, it will say, "Registered users, please choose freely." You take it directly, and once it leaves its sight, it generates a bill. It gives you one or two seconds to initiate payment—one second or two seconds is up to each merchant. If you are a convenience store, think about it: if we have hundreds of thousands of users, and users are shared, and if young people are in a hurry and like to grab and go, won't your convenience store have more customers than the one next door? That's why we launched this product.
These are some of our customers currently deploying in the market, including Gome, etc. The cost is very low. These are some stores we have launched, various kinds. This is a small box ordered by a customer, placed in a high-rise residential building. These are some of our supermarkets. This is our vending robot. The vending robot is used in airports and docks; it's a visual robot. It uses vision to see you sitting there waiting for a flight, and it will come over to chat with you. If you want to buy something, you just take it. If you haven't registered, you can also scan your phone. When you take something and it closes, it deducts money. This product was the first product exported to Australia in April last year. This is our refrigerator exported to Northern Europe; the door flips up.
We think each era should have its own technology. So the future of retail should be like this. We also hope everyone can communicate more with us and experience the retail progress brought by technology. Many people talk about convenience stores. We say change the facilities in convenience stores so young customers can grab and go without going through the checkout. Can you open a completely unmanned store? If you want, you can.
Since DeepBlue returned to China in 2014, we should say DeepBlue is divided into two major parts. The first is our DeepBlue Academy of Sciences. The DeepBlue Academy of Sciences is registered as DeepBlue AI Science Research Center. We have DeepBlue University, which was previously internal and now external. In the industry, our positioning is to be a leading enterprise in the first tier of AI, committed to basic research and application development, with rapid growth, unparalleled market application development capabilities, spanning 7 fields as an AIMaker, with over 20 full-time postdocs and PhD academic leaders, over 100 master's degree R&D personnel from prestigious schools, over 30 AI and algorithm original technologies, over 4 world competition first-place results, over 3 China competition first-place results, over 4 international-level AR joint laboratories, and sales networks and overseas branches in over 17 countries.
After entering the retail field, we have many customers, including those present today and international customers, all cooperating with DeepBlue. We have a basic research team of 70 people. These 70 people have no KPIs; they only do research. Our business products include well-known ones like product dynamic recognition, biometric recognition, text recognition, image recognition, action recognition, posture and gesture recognition, pedestrian and vehicle recognition, license plate recognition, video updates, face recognition, security monitoring, financial bill recognition. For example, ICBC's financial bill system was made by us. We provide autonomous driving technology for many vehicles. Liveness detection and emotion recognition are also possible. If you need, you can contact us. These are the real-time contours, multi-level and multi-vibration collaborative tracking algorithms, and real-time renewal used in our product dynamic recognition. These are also technical aspects.
To summarize today's sharing with you, a few points:
First, we believe retail convenience and vending are trends, and AI will surely take on an important role.
Second, stores should have the capability of being unmanned. That is, let customers grab and go. Whether the store has people or not, it can operate normally. Vending machines must open-door sales, restoring the original shopping scenario; otherwise, customers won't use them.
Third, no matter what kind of change or technological innovation our convenience stores undergo, first and foremost, make it faster for the C-end to buy. At the same time, if it can improve efficiency and reduce costs for the B-end, that's even better.
Fourth, DeepBlue has proposed this phrase, which has been quoted by many companies in the industry: "The C-end doesn't care about the B-end's pain; the C-end doesn't care about the B-end's beauty." Only by making it convenient for the C-end is your retail change correct.
Fifth, products must be able to be recognized at a distance without contact. Machine recognition is the correct direction. We think labeling or other methods will be eliminated.
Sixth, only you can prove yourself. The safest and most convenient payment tool is the human body itself. Before we returned to China, why did the New Zealand government do a lifelong ID project for people? It uses the structure of blood vessels in the palm. Each bifurcation point of the blood vessels is used as a feature value. Infrared is used for three-dimensional reading, and algorithms convert the positional relationships between points into an encrypted number of 250 bits. Theoretically, we can assign an encrypted number to every person on Earth. Once registered, you can enjoy it for life. Whether you are in Shanghai or Chengdu, if you have registered, you can go to the US or Japan and shop with just your hand. No need to exchange money, no need to carry a wallet or phone. Don't forget to bring your hand, and you can shop anywhere in the world. We think this is very cool because you can copy any card or phone, but it's hard to copy a living human hand.
Last, because too many people ask us, "Can we discuss what the future of retail looks like in your eyes? What is the best retail?" We say the best shopping is like taking things at home. Taking things at home is great. Why can you take freely? Because people at home know you. Suppose every store knows you, and every store knows how you will pay. Of course, you can take freely. If you take more, the merchant sells more. So the best shopping is like taking things at home. Technological progress is to make people lazier. AI is to make things simpler.
This is my topic for sharing today. In conclusion, AI changes retail, and the market belongs to enterprises that make good use of AI.
Click Read Original to see more of the 4th FMCG + Internet Conference...
-END-