Kaye: Exploring the Rubik's Cube for Business Growth with Data Thinking | DEV. Together 2021 China Developer Ecological Summit

content source: On June 5, 2021, the 2021 China Developer Ecological Summit hosted by SegmentFault came to a successful conclusion. At the meeting, Netease Smart Enterprise CMO Kaye (Jiang Hanyu) delivered a speech on "Exploring the Cube of Business Growth with Data-Based Thinking", explaining some knowledge of other sectors from a market perspective.

Sharing guest: , CMO of NetEase Smart Enterprise

shorthand compilation and release: SegmentFault Editorial Department

Actually, I happened to be on a business trip in Beijing last week. I have been in Beijing for four or five days, but I remembered the time we shared this time wrong. I thought it was June 15th, but later found out that it was June 5th. A few days later, I flew to Beijing again. Although the process has been a bit setback, I am still very happy to have this opportunity to share and discuss with you. I have listened carefully to the wonderful speeches of several teachers. In contrast, my content is not strictly tied to the developer in a strict sense. You can use it as a different angle to understand something from the perspective of the market The knowledge of other sectors is also welcome to discuss together.

I am now the CMO of NetEase Smart Enterprise, and I am mainly responsible for the overall market work of NetEase Smart Enterprise. Although everyone currently does not have a special understanding of NetEase’s TOB business, NetEase has been expanding its B-side business. Our NetEase Yunxin started in 2015 and has now entered its sixth year. NetEase Yunxin mainly provides IM and audio and video technical services for developers, so this product is closely related to developers. The other two businesses are NetEase Shield and NetEase Cloud Commerce, which are respectively the leading domestic digital content risk control service provider and one-stop business growth service platform.

My personality is more cheerful and outgoing. My hobbies are traveling and running. I am a big fan of Elon Musk. I also like NASA very much. I am always curious about the unknown.

Next, we officially start today's sharing.

I believe that everyone here is no stranger to the term VUCA. VUCA represents the uncertainty and variability of the world. In such an era, I believe that each of us is constantly seeking new knowledge and constantly enriching our own insights. The terms MVP, growth hacker, and AARRR mentioned by the previous guests are actually coming into our perspective more and more frequently in the past two years. In fact, each of us has constantly added new knowledge. I often say to my teammates that in the age of growth, only restlessness can be an excellent marketer. This "restless" growth hacking mindset gives us more possibilities.

I also mentioned just now that I am in charge of the Marketing team. Our team has very important tasks and undertakes the dual responsibilities of "brand building" and "lead introduction". The team started on the first day of its establishment. To put it bluntly, our KPIs are strongly tied to the entire customer acquisition leads, including traffic acquisition and user conversion. The content I share today will start from the actual work of our marketing team, including exploring some of the experience gained during the process of TOB marketing.

When the teacher shared just now, I asked if there are any friends who are involved in the market. I saw many people raise their hands. In fact, the most troublesome or most concerned point for us who are marketers is "Traffic is in where". At present, this picture is also a summary based on the actual work of the team. We have been paying attention to the development of the entire industry. It is obvious that the B2B industry market has maintained a very good growth trend.

Before the speech, I met several VC friends outside the door. I believe everyone has noticed that VCs are paying more and more attention to companies in the ToB field. You may also understand that in the past two years, some TOB companies have been listed in endless news, and the entire industry is showing a good situation. Currently, the overall user habits of users are still mainly search on the PC side. Relatively speaking, the mobile terminal is gradually improving . Many of you here are developers. I believe that everyone will often search for the content they need during working hours. This overall trend is still very obvious. Therefore, the traffic distribution throughout the year is relatively even, mainly concentrated on working hours during workdays. These traffic-related data are extracted from our background and are also based on the observation of Baidu's overall traffic.

In terms of retrieval features, the main focus is on common words, and brand attention is gradually increasing. At the same time, the demand for knowledge of the entire concept is also increasing. In addition, the proportion of after-sales attention has also made us pay more attention to the related launch of the post-sale link. It can also be extended here. Although Yunxin is a technical product, the vocabulary of common words we have put on has reached 150,000. For ToB's business, I think it is still a very large number.

Next, I will talk about the third part. In fact, everyone knows that the entire B2B decision-making cycle is generally longer, which will also cause the entire decision-making chain to be longer, so now brand influence and brand value are becoming more and more important. As a marketer, I can clearly feel everyone's changes in the past two years, especially when attending some summits including technology summits and industry summits. In fact, it can be seen that everyone wants to increase the influence of the entire brand as soon as possible. From the perspective of the media path, the B2B path is relatively long and deep. Despite this, we still attach great importance to content marketing, because it can affect every link in the decision-making chain, so content marketing must be done of. I believe that many of the friends here today are doing developer ecology, and even the predecessors of this industry, everyone can feel that if the company’s assessment requirements for you are, you must learn from the community within three months or a period of time. How many business opportunities are converted, such KPIs must be very stressful for you, but for us, in fact, all forms of content output similar to event salons are collectively referred to as content marketing. At present, knowledge-based platforms also carry a large number of search requests, including domestic Zhihu, Toutiao, etc. Many people have already regarded them as a second search engine. In fact, deploying high-quality content on these platforms is very valuable for the overall impact and transformation of our brand.

As we currently provide cloud services, the characteristics of the population show that most of them are distributed in first- and second-tier cities and developed coastal areas, and users and purchasers are not the same group of people. To deal with this situation is actually what we just mentioned, paying attention to the entire decision-making chain. The different stages of his customer journey determine that the content direction he focuses on is different, even if it is a product for developers. When he chooses models and compares prices, he focuses on different content directions, and these can better guide us to deploy different content.

Now that I have talked about the situation of the entire industry, I will also share what we are actually doing. In China, you have to admit that you want to acquire users, whether it is a short-term user or a company that requires you to recruit new users, we must clearly know the types of traffic, which are the customers that can be obtained by the short-term user, and which ones It is a customer who needs to be nurtured for a long time. We divide the flow into three types of flow, namely bottom flow, basic flow and incremental flow.

The bottom traffic is what we often do now, similar to channels like Baidu, Sogou, 360, Google, and Toutiao. Take NetEase Yunxin as an example. In fact, after launching the developer-oriented service in 2015, we spent a little half a year to construct the bottom traffic, mainly to build some general channels, but also to run through the entire marketing and conversion chain. road. Up to now, the bottom traffic of Yunxin has started to provide some high-conversion traffic continuously and steadily. This is actually what every company must insist on. However, after the business gradually stabilizes and the marketing link runs through, we will pay more attention to basic traffic. The basic traffic will be more similar to the brand effect, and also include some external alliances, such as the Baidu alliance, the Sougou alliance, the 360 alliance, and NetEase Youdao Zhixuan, etc. We will get through all of these. We currently do information flow advertising for the basic traffic segment, such as some brand advertising, which everyone can see intuitively and has a good conversion rate. The third piece is to do incremental traffic. With the continuous vigorous development of the B2B business as a whole industry, especially after 2020, there will be more people grabbing traffic, and the traffic bottleneck will obviously appear. We are also keenly aware that the so-called incremental traffic is actually more high-exposure, but the conversion rate is lower. For example, Baidu launched a small program in 20 years, it is actually more high-exposure, but in terms of conversion rate In fact, it is not so high, but it can achieve high exposure is also very valuable, especially for Baidu system. The Baidu system will support self-built ecology, including what we call know, encyclopedia, Q&A, etc. In fact, the above-mentioned Baidu self-built ecology can even be understood as a free traffic pool, which is worthy of deployment.

Our topic today is "Exploring the Rubik's Cube of Business Growth with Data-Based Thinking". What is more about where our traffic comes from? And how we did it. How should we use data thinking to increase traffic? We know that data tracking is not only the eye of promotion, but also the cornerstone of growth. Everyone should understand some of the work we are doing now. It was just mentioned that we have 150,000 general vocabulary, and we will add some parameters and hyperlinks to every promotion. I think this is very important. .

For example, we held a 1,000-person conference in Beijing last week. All registration data, attendance data, and WeChat group data are all QR codes with parameters. For us, all traffic tracking is actually considered. We have always been paying attention to where the traffic will end up and how the conversion results are. Whether it is some activities or promotion and operation, data tracking is required. As for the specific way to do it, it depends on the project situation. Because some channels like LinkTag do not allow you to set parameters, but the tracking methods are still relatively diverse. Many of our NetEase data burying points use our own set of self-developed systems, called Netease Hubble, through burying points Can track users' habits more comprehensively. After building our channels, there is a very complete evaluation system. At present, we will directly track the completed order after all the traffic is introduced, and it can be synchronized to the backend or the database in UBA or CRM when the user registers, so that we can know the source of each clue and evaluate the channel effect.

After we introduce all the traffic, we will directly track the completed order. In this process, we will evaluate through three dimensions: basic indicators, product indicators, and profit indicators. When it comes to building channels, the first thing is to set up basic indicators to better evaluate the scale of the channel. What is the effect of drainage? How many exposures? How many clicks are there? In terms of product indicators, we will evaluate the quality of the entire channel, and the bounce rate of core pages? How long is the average visit time per capita? At the level of revenue indicators, we will directly track whether the traffic I attract can finally be completed? Can it bring performance to the company?

Our landing pages will also be divided into several types, including functional landing pages and basic traffic landing pages. We also pay attention to channel quality. For example, some core pages seem to have a lot of traffic, but their bounce rate is very high, so we have to consider whether there is a problem with the content layout, and whether the content such as the introduction and the content The keywords did not match the information received after the user arrived on the page, which caused the user to jump out. We all have relevant personnel to pay attention to and follow up on these issues. From the perspective of growth, if the page carries 1,000 visits, but in the end only 10 is converted, from a certain dimension, we have a lot of room for growth. From the perspective of input and output, we now have a formula for calculating market input and the final result of the order: LTV / CAC. LTV mainly refers to the revenue that a customer contributes after becoming your paying customer, and CAC refers to all the costs we have paid to obtain traffic. There are some differences in cost calculations, but marketing expenses, marketing personnel, and sales labor costs are often included. At present, we think that if the number can reach about 3 after this, it means that the quality of the channel is still very good. If it is greater than 3, it means that the channel is very good, but if it is lower than 1, then say The truth is a bit like a losing business.

We now have five sub-products under our entire product. Under better circumstances, we can achieve this number from 5 to 6.

I have just introduced so much. In fact, I want to tell everyone that we believe that the data operation or digital marketing of TOB is essentially to use data analysis to drive growth. I think the concept of data-driven is still deeply rooted in our team. We think that the entire operation is actually a continuous process, not what should be done or not, nor is it driven by experience, but it should be driven by data.

The second more important thing is the familiar ABtest. Abtest is relatively easy to land for some high-traffic websites on the C-side. Sometimes you change the registration button to a guideline, and the traffic conversion can be more obvious. But for ToB, the traffic is actually much smaller. , Many times when we promote ABtest, we will be questioned about his input-output ratio. But I think no matter what, the verification and testing of everything actually requires the ability to truly understand it in a different way. I think it is necessary to firmly know about ABtest, and he can really use data to drive growth.

The third piece is to develop data indicators. In fact, our company now has a lot of business, and in the early stage, each sub-product business was developed relatively independently, and there were different sales teams, sales directors, and different product teams. The marketing team is the only team led by me. In the process, you will find that people often have different language systems. In this business, he likes to call registered customers, and that business likes to call Leads. If you change another business, he said he prefers to call effective customers, and another business is also called transfer customers. In fact, this will not be conducive to the overall data connection in the later period. This seems to be a very simple matter, but the same language system and the development of the same data index system are still very important to me. Our marketing team, data team, and sales team all talk about conversion rates. Don't argue or blame the other party. Why don't you follow up? It's obvious that this works, this doesn't work, and so on. We will all pull out the data. I also wrote above that there is an SDR team, and I will tell you more about it later.

The first part of the previous section mainly introduced you to where the ToB traffic is, what we need to do based on these traffic, and how to analyze the value of these traffic.

In the second part, I will briefly share with you the closed-loop marketing data to improve conversion and promote growth. In particular, I mentioned the SDR team to everyone earlier.

This picture was made by our team and the data team together, because we believe that the entire growth routine is generally: through combing our own business situation, using data to describe the key indicators of each link of AARRR, and finally determining the North Star of each growth project index. After more than a year of experimentation. We actually abstracted a set of growth solutions for ToB. All the things on this picture are researched by ourselves, and we will do some data burying and so on.

see that the top layer of this picture is the customer acquisition that I introduced to you in the first part, including market, sales and data. This picture shows that the traffic is at the top level. In the first part, I mainly introduced the traffic from search engines, but in fact, the sources of our traffic are very, very wide, including search engines. Similar to today’s offline conferences, it also includes our online activities, as well as some in-depth content marketing that we often do. Sometimes you can spend a small amount of money to split a set of operation growth manuals or obtain white papers. In fact, they have achieved good results. NetEase Yunxin is mainly for developers, so we have translated a set of audio and video technical standards, called WebRTC standards. Until now, we are still bringing us a steady stream of potential leads. Currently, we have brought nearly 2,000 leads. The content we do is almost free for developers, but our content staff will meet in the early stage. Spend a lot of time to do some planning, translation, content arrangement and channel deployment. It can be clearly seen from this picture that the top layer is the source of our entire traffic.

second level of 160ebbe4b8b02d is conversion, and the clues are filtered from MQL to SQL. the corresponding data tags, we extracted the rating model of the entire lead and the evaluation model of sales force. These two models will also be briefly introduced to you later. The rating evaluation model of sales force is mainly based on the customers who are good at sales and the characteristics of the clue itself to determine which sales the clue best matches. The last level of this picture is retention, which mainly includes sales and customer success. We can find that the traffic has been followed up from the top to the bottom. If the customer is lost, then our entire data analysis includes subsequent purchases and additional purchases. The corresponding customer service must be adjusted accordingly. In fact, since the launch of our marketing department, the data burial point has existed. From the user's interaction, such as the user to register or browse, until he is completely lost, or it may become my customer for me to follow up. Service, he is in our entire TOB data growth model.

We just talked about the circulation problem. In fact, the B2B clue circulation path still has a relatively ideal closed-loop traffic. Regarding the closed-loop traffic, I think it can be summarized in a few simple sentences. Expanding the traffic funnel, registering timely response, efficiently assigning clues, releasing sales energy, cultivating clue systems, and optimizing market drainage can improve our customer acquisition efficiency as a whole. In this whole link, we will actually see a very deep cooperation between market and sales.

Regarding the closed-loop practice of B2B marketing data, I think it’s just like what I just said, everyone must have the same thinking and language system for many cognitions, and we call it Dao and Shu.

so-called road of 160ebbe4b8b12b is that we must be customer-centric and become a highly coordinated growth flywheel. Speaking of this, everyone will feel that the customer-centric words seem to be quite imaginary. But I think that in many cases, especially in actual work, there are some core concepts or some culture. It still needs to be delivered continuously. In fact, under the cornerstone of customer-centricity, we require a unified set of speech system, unified goals and strategies, unified systems and technologies, because we have just mentioned that our entire product business has Different sales teams sometimes find that there are some communication obstacles between different product teams in meetings. I believe that if you have so many kinds of businesses, so many kinds of products and product teams in the company, this The feeling is still obvious.

Just give you a brief introduction of the road, but in B2B marketing is actually more practical , that is, then I want to share with the surgery, there will be three parts to share. The first part is what I just talked about. To get through marketing data and track user status to the end, uses data to speak . The second part is the thread circulation strategy . We will continue to carry out refined thread circulation link design based on the actual situation. The third part is SDR talent management . The SDR team was introduced by me in 2018, and it was officially established in 2019. Now to 2021, our entire SDR team has undertaken almost all the lead traffic of NetEase intelligent enterprises. They will release some of their sales energy, which can greatly To improve the conversion rate.

first part of 160ebbe4b8b1e3 is to use data to speak , which can be divided into three small parts. First, the source of Leads can be tracked, and user behavior can be analyzed. You can see that all of our data can finally be associated with an ID.

Due to the characteristics of China's national conditions, most of us now use mobile phone numbers as unique IDs. Therefore, it is very important that the source of Leads is traceable and the corresponding user behavior can be analyzed. Our user behavior analysis tools are self-developed, and our customer relationship management system is also self-developed. We can aggregate all customer data under one ID, and aggregate all user paths together.

The source of the channel is written on this sample picture. These are actually all of our promotions that I talked about earlier. Whether it’s activities, content or search engines, they will have parameters and channel numbers. It can help us to conduct better data analysis in the follow-up, including the source of the page and the user's most recent visit time, so that we can understand the user's behavior path more comprehensively. On the right, you can see the information that users fill in when reading our articles, and all the data that users fill out actively will also be recorded in our system. With data as the cornerstone, we have also built a marketing management cockpit. In fact, we also hope that there will be a full-link data board from market to sales to service. We will not talk about other things when we hold weekly meetings every week. You will look at the data first. How does the data perform? From the data board, you can clearly understand which link has gone wrong and which part of the conversion has dropped.

second point is to standardize the filling of CRM, and do a good job of data precipitation and governance. I am quite aware of this. I remember that before the establishment of the SDR team, we wanted to analyze why the registration data is invalid. Everyone knows that invalid data will be thrown into the high seas pool. We will find out after salvaging the data. There are more than 3000 reasons for invalidity. In fact, the reason is very simple. Since everyone fills in the information randomly, many data cannot be structured and analyzed. The exported data is messy, so after I set up the SDR team, I asked us to be at every stage It must be filled in very standardized, so as to make the subsequent data analysis more accurate and clear.

After the previous part is implemented, it is very important to carry out a refined clue circulation link design. The left is a picture of a foreign country, and the right is the corresponding lead circulation process of our team. In fact, the general idea is the process of lead from flow to market approval to sales approval, order completion or loss, and finally the process of re-operation. The design of the entire link is very meticulous, which requires both market students, SDR students, or sales students to be very clear about what I should do and what I should pay attention to at different stages, such as , I’m now communicating with our sales director. At this time, my data has met certain conditions, and it can be transferred to sales for follow-up. When sales think this is an invalid lead, then I will do the same Whether the review is due to not satisfying some prerequisites or is it due to the personal judgment of the sales.

I just told you that we have a system for rating the entire leads. When users come to our platform through online registration or other methods, their data will not be directly given to the sales or marketing department. The data will also go through our set of clue rating system to automatically calculate his score, which is divided into three levels: L1, L2, and L3. Then everyone will definitely ask how to score it? We actually have a feature weighted scoring model. This model is mainly led by the marketing department. We will work with the sales team to make some regular model modifications. The model is Lead Score = W1 F1 + W2 F2 + W3 F3 + Wi Fi, where Wi is the weight of each feature, and Fi represents the feature. For example, if the registration information filled in by the user includes a complete business name, a complete personal name, and a complete mobile phone number, then he has the corresponding Feature, his weight will be higher. When I talked about the characteristics of the population, I also mentioned that most of our current clues are from first-tier and second-tier cities. If the registered clues are from Beijing or Guangdong, then its weight will be relatively higher. One point, for example, if the user’s company is a Top 100 in the Top industry, its weight will be higher. In addition, because we are able to capture registration data, which pages he browses or even downloads our demos before the official registration, once he has these interactive behaviors, it will increase his weight. As for the specific weight of each W, we will adjust our standards according to different stages and conversion rates. We have been doing this for nearly six years now, so we have a huge amount of user data. Therefore, our SDR team will also layer user data according to the scoring model when conducting data follow-up and data salvage.

I have introduced the features of our scoring model to you before. Here I will show you in detail. The features of our Lead Scoring scoring model include official website behavior, interactive information, company information, product usage and registration information, all of which will actually become our score. The characteristics of the model.

Since each company has its own characteristics and different actual situations, if your company wants to build a similar scoring model, we will pull out the past transaction data before doing this model test. See, to analyze which characteristics of the clues are relatively higher conversion rate and transaction rate, and then abstract the corresponding data to form the final model. And our data team will also dock interfaces similar to Tianyancha, because we, as the marketing department, are familiar with some well-known products. For example, I will clearly know that Douyin is a product of Bytedance, but this is for sales staff. It may be unfamiliar. By docking a system similar to Tianyan Check, it can help us identify whether the registered company is Top-level in the first time.

I just mentioned that our marketing department and sales department will clean data through SDR. Only after registered users meet the lead standards can we successfully pass our cleansing. For the selection criteria of MQL, BANT is one of the more common ones on the market. B stands for budget and A stands for influence. This actually involves the language of our SDR team. When team members go out to call, they can pass the standard language. To ask whether the user has sufficient budget and whether the registrant of his company is a person with decision-making power and certain influence. The third point is to ask whether his product needs can be met by the service we provide. Fourth One is to ask about the specific project start time. They may want to do a project, but they are still at the stage where they want to do it. The real project will start half a year later. For example, we will try to filter out those who are in the selection stage and are already in the selection stage. Promoting projects, allowing sales to follow up quickly. When these clues were initially screened according to the standard, our SDR played a very important role.

I have been talking about the term SDR to everyone just now, and some students wondered what exactly is SDR? In fact, the Chinese name of SDR is Sales Development Representative. We also surveyed many companies when we formally established the department. SDR currently has two types of responsibilities and division of labor in the market, but different companies may have subtle differences, but in general they are divided into these two types. The first category is mainly to clean the market through outbound calls, obtain clues through market cleansing, and classify the clues. In fact, this type of SDR is more to do some passive basic outbound work, and does not need to expand out by yourself. However, the other type is B-type SDRs that need to take the initiative to make outbound calls. They have to actively expand like sales. Such SDRs can get a commission. Currently, companies on the market will give two points to this type of SDR. About the commission. The first type of SDR he takes is the basic salary plus the assessment salary. For example, our company is still mainly class A, and my SDR team mainly operates with a more standardized process. The work process of SDR can be represented by such a diagram, including the introduction of market leads to sales follow-up to customer transactions.

first cleans the data of potential customers . If it is a SQL intended user, it will be transferred to the sales follow-up. If the sales can be transacted after the follow-up, it will be our final SO service. If it is not intended for SQL users, they will enter our aquarium, and we will also cultivate them again. In this whole process, it will be combined with the data model rating we mentioned earlier. In fact, before the SDR outbound call, the clues are already rated. After the rating is completed, all of his behavior data to the final transaction, we are all Follow up continuously in the form of an ID.

Why do you need an SDR team? In fact, everyone knows that although we all emphasize the importance of data labeling and CRM follow-up needs to be timely, sales in many companies are often just sales because they are evaluated by strong KPIs. For sales, they are more concerned about how to get the order. As for whether the remarks in the follow-up process are clear, whether the follow-up is timely, and whether the label is complete, these are not their primary concerns. Under such circumstances, our SDR, as the first round of data clue screening, can better improve the data, and do a good job of data annotation, including data feedback. This actually goes back to what we said in the first part. When we do delivery, we need very accurate data feedback. This is mainly due to the collaboration between our SDR and data delivery personnel and complement each other to complete the work. At the same time, they are in the data Promotion, conversion and cost saving also play an important role.

In terms of cost saving, I myself have some practical experience. When I launched SDR in the first year, our sales director actually opposed it. This is normal, because all the clues and data before are on the sales side. They are also accustomed to holding clues in their own hands. They think which clues should be in charge of sales, and they are unwilling to take the clues out. In this context, SDR is indeed in the first year of advancement. Difficulties. So how should we break through? How to settle this matter? So in the first year, our SDR team only recruited three members. The first requirement is to make increments. The so-called increment is to build a data model from the data discarded by sales. I also mentioned that the sales will be correct. The clues they say are useless are thrown into the high seas pool, and we just go to the high seas pool to make data models. Salvage the data that we think can be salvaged first, for example, a certain batch of data is not satisfied with a certain function within three months, or because of the price system and other reasons. There are also clues to our activities, because many companies’ sales are not interested in following up on the clues of activities. I believe that there are so many people in our Developer Ecological Conference today. If your company is doing developer ecological products, then You give these registration data directly to your sales, most of the sales are unwilling to follow up, because he would think that this kind of data is extremely inefficient, so our SDR started from here at that time. In fact, three SDRs are really not expensive after counting. Just five or six thousand yuan of salary plus social security, each person will be less than 100,000 a year, and three people will count as 300,000. But it was through this kind of high seas data salvage and other clues that we made a profit of 2.6 million. You can understand that I spent 300,000 labor costs in exchange for 2.6 million in revenue. This is indeed a good deal. Sale. After this incident, it also proved that SDR has great value in terms of data label improvement and cost saving. Gradually, SDR has begun to take over the data clues of our entire NetEase intelligent enterprise. I think that many times, we always think that data is a very tall thing, but I think that data actually exists in the bits and pieces of daily work. We will feel that data should be a set of very large systems, a set of very deep theories, and various models are needed. I think these are necessary, but it is also important that the early release of the information, the improvement of the data label and the daily follow-up of the data label are timely and accurate. Only when these bits and pieces are integrated, can we really do a good job in digitization. drive. This is also our true feeling after several years of actual combat.

Today is the developer's conference. I also said before sharing. A lot of the content I talked about seems to be not very closely related to development. I also told the organizer that I will add a meal to everyone at the end and tell everyone about it. Twilio this case.

I believe that the classmates who are developers have more or less understanding of Twilio products. These materials are 3 pages in total. This is actually a sharing made by Twilio’s investor BVP. Twilio is a developer-oriented product that aims to provide developers with all convenient tools. It was established in 2008. Listed in 2017. BVP started investing in the early days of Twilio, and later led the round B, C and D rounds. After Twilio went public, he has grown from the earliest US$50 million to the current 57 billion. The entire growth has actually turned over. 1140 times. I believe that its growth rate or rate of return must be very envious of everyone. As we will see, in fact, Twilio is a very typical company that pays great attention to the developer ecology and the bottom-up communication of the entire developer.

In fact, I think he emphasized the third point very much. For a long time, they actually paid great attention to organic growth and the spread of developers from the bottom up. Many products on the market, including foreign Slack or Zoom, are based on self-propagation systems. Twilio will also organize some activities in the early stage, including content marketing, etc., but they still firmly believe that only through developers and realizing developer self-propagation can they truly break the market.

On the last page, BVP mentioned that he is not so famous in China, but he is actually a very old investment company in North America. During the sharing, they made eight suggestions for the developer platform. They also believe that a developer platform with refined operations can provide very good assistance and blessings for sales. At the same time, if the entire developer ecosystem can be built well, it can actually give full play to the natural value-added trigger mechanism inherent in engineer resources and products. from. The evangelism of developers can also promote product use and sales. The eight suggestions he gave, in general, still emphasize that we must build a developer platform to better update and iterate our products, so that our cognition and engineering resources can be optimized and used to the greatest extent. In fact, I think NetEase is just starting out as a developer ecosystem, and even I don’t think I’m fully started. I’m not here today to share with you. It’s more to communicate and to learn, just like I just said. For the overseas cases I visited, I am also learning some advanced overseas experience. I also hope that today's speech will be helpful to everyone. Thank you.