domeet webmaster

When technology and data become new production factors, What changes will China’s residential industry face? Recently, I interviewed Yan Mi, the CTO of Shell Finding Housing, and talked with him about the technical thinking of shells and how technology can reshape the new Residential industry < span>.

In this interview, Yan Mi shared four judgments:

1. Data is more important than algorithms. If the data foundation is good, AI engineers will become “adjustment engineers” in the future;

2. SaaS has changed from 1.0 in the PC era to 2.0 on the mobile side, and will be SaaS 3.0 based on AIoT device context awareness in the future;

3. Service industry AI cannot replacePeople, whose core is to solve the problem of variance among practitioners, there will be great opportunities for AI assistants in the field of toB in the future;

4. In the future, hardware devices will become more versatile. Just like smart phones, industry applications can be written by APPs and software. In the future, IoT platforms for more industries will appear, making the combination of software and hardware easier.
   

   Here is the conversation (edited).

   

   Shell find room CTO Yan Mi

   Talking about the shell: Internet practice in the residential industry

   : Shell positioning “technology-driven new residential service platform”, what dimensions does technology here specifically include?

   Yan Mi: The technology of shells is to build a brand-new data interaction around people, rooms and guests.

   On the one hand, we have three underlying data models: one is the knowledge map based on the real estate dictionary, the other is the relationship map of people, rooms, and customers, and the third is the server model, which is used to classify brokers. Credit grading to define professional broker portraits.

   On the other hand, we are also innovating in intelligent scene applications, such as Beike intelligent hardware and Beckham AI assistant.

   : What is the thinking of shells when using technology to transform the residential industry?

   Yan Mi: The first step is the most basic step, which is to manage the business processes and data first through the system to achieve onlineization and standardization. The second step is data mining. With data, intelligence can be done.

   After Tencent put forward the concept of the Industrial Internet in 2018, now everyone is talking about informatization, standardization, and intelligence, but the shell has indeed explored a lot, which is also in line with it.

   Why does the Industrial Internet appear? You have to look at the development of China as a whole, because every industry is faced with the problem of improving efficiency-the disappearance of user bonuses, higher costs, and increasing competition. The Industrial Internet is to digitalize all aspects of the industry, and then analyze and optimize it.

   : How did these shells go?

   Yan Mi: Chain Family began to digitize and standardize people and houses in 2009, and it has done so in the past few years. There are two things worth noting about digitization:

   First of all, you can’t just look at the magnitude, but also the depth of the number. What the user sees is a listing on the shell, but there are 443 fields corresponding to the back of the platform, which has become the most granular and widely covered housing information database in China. , Built up little by little;

   Secondly, it is necessary to make the closed loop of data acquisition natural in the business process. The information in the offline link is manually entered, and the conversion will be missing. For example, the agent showed him, who did he take, and which house did he see? Did you really go? Are you looking at the real house? This information is not completely guaranteed to be true and effective. This is a problem to be solved in the future. Smart devices can be used to make data acquisition more natural.

   And intelligence is a natural thing. A year ago we suddenly found that there was a lot of data in our hands. Is it possible to use a mature algorithm to do a little AI? We have made a few good applications now, such as the uneven service level of brokers, we have designed some tools, to make the variance between brokers smaller.

   For example, Beckham ’s assistant, Some brokers do n’t know how to answer customers ’questions, or even how to ask, Beckham ’s assistant can analyze how a senior broker does it, and then prompt him how to answer . When a user asks, is there a lift in this house? Little Bei will prompt: Do you have elderly or children in your family?

   : How many smart hardwares does Shell have now? What are their characteristics?

   Yan Mi: We established the Beike Intelligent Hardware Team in July 2019, and released the Smart Bracelet andSmart door lock.

   The bracelet and headset we designed are integrated. The headset is placed in the bracelet for easy access. Because the broker is most concerned about business opportunity news, we have made a customized process. The news reinforces the shock.

   Our smart door lock is a sticker lock. It used to take 2-3 hours to change a lock. It only takes 5 minutes to stick the lock. It is good for temporary use. And it is suitable for 80% of domestic listings.

   : Will these smart hardware consider opening up to scenarios outside the platform?

   Yan Mi: I hope to be able to open it, like someone already asked about the door lock. But the device itself is not mature yet and will not be considered for the time being.

   

   Thinking: AI in the service industry cannot replace people

   : Do you have any thoughts in the process of exploring the Industrial Internet?

   Yan Mi: Let me make a few points.

   First, data is more important than algorithms. If the data foundation is good, AI engineers will become assistant engineers in the future-models and algorithms are readily available, and tuning parameters is fine.

   As said above, we suddenly found that there was a lot of data, and then we had the next step. For example, to make a VR model of a house, we used a big camera and 9 cameras to scan the house. But after making 1 million sets, we found that we can do this with deep learning: you can take a normal photo with your phone and you can pass deep learning estimates With data such as depth, equipment costs are reduced at a glance.

   So you can do a lot of things after you have data. This is a typical feature of the Industrial Internet, because you only have data in the industry.

   Second is the 1.0 of SaaS from the PC era.The exhibition will be 2.0 on the mobile side, and the future will be SaaS 3.0 based on AIoT device context awareness.

   With the development of smart phones, SaaS has been moved to mobile phones. It is regarded as the 2.0 era, and business personnel can operate on mobile phones, such as logistics, express delivery, catering, and hotels. And we have realized mobile SaaS as early as 2015.

   But I think that SaaS in the 2.0 era is just a recorder and dispatcher. After the business staff has done something, they open the mobile phone and enter it. But how do you ensure that his entry is timely and true? The system distributes tasks to business people, but how to ensure that they are in place?

   We have designed a lot of mechanisms, processes, cross-validation, and invested a lot of resources to do operations to ensure the authenticity and reliability of the data, but the process is actually painful. If you ca n’t guarantee the timeliness and authenticity of the data, it ’s useless to make the system. I dare say that the authenticity of the data in most SaaS on the market is not more than 50%. Source data, the true rate does not exceed 30%.

   I hope to enter the 3.0 era as soon as possible. SaaS will rely on IoT and AI when collecting data, and rely on the device’s own perception to make data acquisition more real and natural.

   Third, many people are saying that AI will replace people in the future, such as industry and manufacturing, but I think that the service industry is different. AI cannot replace people. The core is to solve the problem of variance between practitioners.

   The service level of the brokerage industry is uneven, some are small and others have rich experience. If you set excellent people as a benchmark, dig out how they do it, and abstract some rules and models, they can be used to guide people who perform poorly and raise them to an average level.

   The machine cannot be better than the good people, but it can be higher than the average level. The AI ​​lecture room we made last year has actually exceeded the average level of people. Its average listening time and listening completion rate are much higher than the brokers. People mean. But this year, we will spend more energy to try to help ordinary brokers improve the level of lectures through machines, instead of directly replacing the agents to lecture.

   For service industries including sales, there will be great opportunities for AI assistants in the future. A few years ago, a group of entrepreneurs with NLP backgrounds were thinking about the opportunities of various AI assistants and voice assistants, but they were all toC scenarios, obviously not satisfactory. But toB’s AI assistant scene will come faster and more violently. toB has a knowledge base, training, performance evaluation, strong elimination, and mysterious customers and NPS. In short, there is work pressure, and toC has nothing. What does AI do?

   We will launch the “Beckham Assistant” system in 2019The listed products are the typical toB AI assistants, which can help brokers chat, intelligently dispatch brokers to reach customers, and so on. I hope that they can be carried forward in the future.

   Fourthly, hardware devices will become more universal in the future. Just like smart phones, industry applications can be written by APPs and software. In the future, IoT platforms for more industries will appear, making the combination of software and hardware easier.

   Combining software and hardware is a very difficult thing. We communicate with many hardware manufacturers. I put forward requirements. They do n’t know how to do it, or they are not willing to do it because of difficult R & D and long R & D cycles. In the past, many industries wanted to do bracelet projects, but the combination of software and hardware was really difficult. In the end, 95% of the projects were dead, and hardware manufacturers were reluctant to do it. And I made a lot of hardware pits from scratch, and it is difficult to produce good products after a few years of polishing. However, I believe that hardware devices will become more and more versatile in the future, just like smartphones, and only need to develop software, such as bracelets, watches, headphones, and so on.

   In the past, we said that the IoT platform is in the fields of smart home, agriculture, and industry. I think there will be more IoT platforms in the toB field in the future. As a middle layer, this platform connects devices and industry SaaS at the same time. It abstracts SaaS functions and hardware capabilities, which makes it easier to combine software and hardware. This is a typical PaaS. In the future, there will be more and more universal PaaS platforms for certain industries.

   

   In June 2019, the Shell CTO line identified four important digital construction directions in the future, and IoT business is one of them.

   Talking about industry: using technology to assist in establishing professional attributes of the industry

   : Compared with foreign countries, what is the degree of digitalization of China’s real estate industry at this stage?

   Yan Mi: Much higher than the United States. We have studied the United States for a long time. It can be said that the United States has a very prosperous ecology. In contrast, it is probably the only company in the country that has enough resources and money to do this.

   For example, a real estate dictionary, which has been done from the chain home to the shell,Now the number of real listings has exceeded 206 million, covering 500,000 communities in 326 cities. It is now China’s most extensive and granular listing database.

   There is also the data of RTV, which now has 3.3 million sets of VR housing data, which is already the world’s largest VR database.

   : But there is a complete public housing database in foreign countries.

   Yan Mi: We hope to learn from the US MLS model and require platform merchants to share and syndicate, but we can only look at this from the perspective of property cooperation. . The penetration rate depends on how many houses are sold in shells, which vary from city to city. The lowest is only 20-30%, and the most can reach 60% or even 80%.

   : If you want to achieve a leapfrog development in the field of residential services in China, what do you think are the driving forces or potential points?

   Yan Mi: This matter can be viewed from two levels. The first is the level of intelligence-space intelligence and service intelligence. In addition to the three-dimensional data of the listing itself, the intelligence of the room is also reflected in the decoration and other aspects of the space; what can be done in the space of the store? Chain households need tens of millions of electricity charges each year, and intelligent management of energy is also an aspect; intelligent service, the core is technology to reduce the variance of service providers.

   There is also a talent level, to judge whether a industry is mature, you need to see if it has professional attributes and whether it has a professional training and growth system. Many employees in the service industry are just for a living and are unwilling to develop. This cannot be called a profession.

   The role of that technology is to see if it can help build such a system. Shells work hard to make many cities brokers’ careers from six months to one and a half to two years. However, The most fundamental thing is to use technology to improve the efficiency of the broker and make money.