Deals | Tapping into The Unmanned Retail Business with Low-Cost Solutions, Haha Store Banks Tens of Millions of RMB

The startup seeks to trim cost while embracing opportunities in new retail.

By on

Deals | Tapping into The Unmanned Retail Business with Low-Cost Solutions, Haha Store Banks Tens of Millions of RMB

Kr-Asia is all about actionable insights for entrepreneurs. And through this post, you’re about to find out:

  • How to trim cost while embracing new technologies?

  • What are the merits and demerits of the three technologies that the unmanned retail business relies on, i.e. image recognition, RFID and palm vein recognition?

When Amazon Go made its debut, everyone felt that the future of unmanned store is here, just “grab and go”. Over the last year (2017), China has witnessed the emergence of many different variants of unmanned stores. There are, for example, RFID-enabled vending machines now dominant on the market and the visual recognition technology-enabled ones that birthed only until recently.

Haha Store (哈哈零兽), which develops and supplies vending machines enabled by image recognition technology, has just netted Pre-A round worth tens of millions of RMB. The round is a co-investment by Pine Capital Partners and Yuanwang Capital. And the capital will mainly be used for development of algorithms and marketing.

Founded in Wuhan, Hubei in early 2017, Haha Store (formerly known as Haha Bianli) specializes primarily in providing retailers with smart vending solutions, including vending machines, algorithms, retail software solutions, e-payment portal (authorized by Alipay and WeChat Pay to process payment without password) and more. Haha Store has, as of now, manufactured hundreds of frozen food vending machines. And it had just delivered a batch of vending machines to the snack brand Bestore and the fruit retailer Pagoda three months ago. Additionally, the orders for its vending machines keep flowing in, totaling over a thousand vending machines from 20-plus companies.

Here’s how its smart vending machines work. For starters, the shoppers have to scan a code to open the door. As soon as the shoppers get the desired items and close the door, the camera installed inside will take a snapshot. Then, the algorithm will figure out automatically which items the shoppers take and charge the shoppers’ WeChat Pay or Alipay accounts accordingly. The vending machines of Haha Store are, currently, capable of identifying nearly 80% of daily products, including bottled water, snacks with rigid packages and fruits, with an accuracy rate reaching 99.05%. It’s also worth noting that the algorithm can be further refined through complete self-learning.

Retrofit, RFID and palm vein recognition

Image credit to Haha Store.

Well, if you dig deeper, you will find that the vending machines of Haha Store are actually all retrofits of ordinary freezers. The Chinese market sees no shortage of freezers, while getting a new mould could cost millions of RMB. So, Haha Store has technically saved millions of RMB by doing so. Anyhow, the freezer could be converted into a smart vending machine by adding the following parts: anti-theft lock, several cameras, sensor and 4G network modules. “This solution has helped bring the unit price of Haha Store’s vending machine down to only a few thousands RMB, and that’s even before it is thrown into mass production. So, it is very appealing to retailers,” FAN Wei, founder and CEO of Haha Store, told KrASIA. As a matter of fact, there are another two technologies on the market that can achieve the “grab and go” shopping experience. But, they both come at high prices.

One is RFID. The antenna needed for an RFID-enabled vending machine costs dearly. Furthermore, antimagnetic treatment on the front door is needed if the vending machine is to work properly. Besides, the vending machine can only be assembled manually, which makes impossible its mass production. So it would come as no surprise that its price is so high. An RFID-enabled vending machine now costs at least ¥10,000, not to mention that each product shall be labeled with a tag which cost RMB0.3-0.5 each. The operational cost is just too high.

The other is palm vein recognition. The AI vending machine by DeepBlue Technology, for example, is based on this technology. With this technology, shoppers can unlock the front door simply by scanning their palms, which bear unique biological characteristics. And the check-out process (items calculation and charging) of the vending machine is a consolidation of both image recognition and palm vein recognition technologies. This will undoubtedly add to the total cost of the vending machine.

Edge in the market:

Image credit to 123rf.com.cn.

According to FAN, Haha Store has a leg up on those potential rivals in terms of the following two aspects:

First, it is four to six months ahead in R&D of its vending machine. Stability, citing Fan’s words, is key to hardware products. They have to be tested in all kinds of real application scenarios and accordingly upgraded.Haha Store’s vending machine has since its trial operation in September 2017 gone through 5 months’ test. Its vending machines are now fit to roll out and be manufactured at a large scale.

Second, its algorithm has become much more refined over time. Haha Store’s AI system is now capable of correctly identifying around 1,000 products. This is really a tough to follow, as training and refining algorithms take time.

Team:

Haha Store now has 40 staff. Its founder Fan had worked in Tencent as product manager and as business development manager in group-buying platform Meituan while Meituan was still in its infancy. Additionally, Fan is also a serial entrepreneur who had previously started businesses in hardware products, food materials B2B business and catering. His past experience has shaped his in-depth understanding of software and hardware development, its mass production, procurement of retail products, storage, site selection and operation. Key members of Haha Store are all former employees of Baidu or Huawei. And its AI researchers have all been working with objects identification for over 10 years.

Writer: HUANG Xuejiao

Editor: LU Zong di, ZHAO Xiaochun