When Kuaishou Technology decided to enter the toB business, it did not want to repeat the industrial cycle of high investment, heavy manpower and low gross profit, which was an exploration of new business for Kuaishou Technology himself; for the toB industry, it might be a reshaping of the business model.

文｜Jonas.

Zhihu Inc. 's pain point

At the beginning and end of 2020, Zhihu Inc., the largest question-and-answer community in China, found Kuaishou Technology. In that year, it was the key node for Zhihu Inc. to seek media upgrading. In October of that year, Zhihu Inc. announced that he would spend 500 million yuan to encourage video creators. Determination and funding can be put in place quickly, but the foundation of technology needs to be more solid. At that time, Zhihu Inc., who started as a picture and text community, needed to accelerate and strengthen the support of video technology. Take "video coding", a crucial compression technology in the video field, as an example. Zhihu Inc. used H.264 technology at that time, while head video content platforms such as Kuaishou Technology have already used the more advanced self-developed KVC codec standard, which saves 30 to 50 percent of the bandwidth than H.265. The technical difference between the two generations means that by compressing videos with the same definition, Kuaishou Technology can compress files smaller, make users look more silky, and save a lot of bandwidth and storage space fees without taking up a lot of bandwidth. "clarity" and "fluency" are the most solid cores of Kuaishou Technology in providing audio and video services to the outside world. these two indicators seem to be basic, but in fact, the clearer the video, the higher the cost (because it takes up more bandwidth). The more fluent the video, the lower the cost (because the small code rate of the file is low). In other words, clarity and fluency, the two elements that need to be possessed at the same time, have contradictory characteristics, which requires Kuaishou Technology to find a delicate balance through data-driven, and on the premise of ensuring a clearer video, it can still play smoothly on the client side, and the cost can be controlled. But how to find this critical balance? This is still a strange technical blind spot for Zhihu Inc., but it is Kuaishou Technology's usual prescribed action. Audio and video and AI are the base of Kuaishou Technology's business. Kuaishou Technology has more than 300 million daily active users and nearly 600 million monthly active users, and the amount of video he processes every day is very large. Video playback VV is of the order of tens of billions, and live streaming is of the order of hundreds of thousands. Many technologies have been honed in the whole link of coding, compression, transmission, playback, distribution and so on. Zhihu Inc. dares to hand over the olive branch of toB cooperation to Kuaishou Technology, a standard toB company, a bigger reason is the trust in the fast hand audio and video team. Since joining Kuaishou Technology in 2016, Yu Bing has been looking for ways to attract technology heroes in the audio and video field, and has spent six years building a luxury team covering live streaming architecture, short video architecture, video codec, transmission algorithms, audio algorithms, etc. The cooperation with Zhihu Inc. was an important exploration in the early stage of Kuaishou Technology's toB business. In nearly two years of verification, Kuaishou Technology also opened cloud service cooperation in the audio and video field with China Unicom, XIAOMI, CCTV, Duoduo Video and other manufacturers. In the process of creating continuous business with front-line customers, internally, Kuaishou Technology has a clearer understanding and confidence of self-research technology; externally, Kuaishou Technology has a more accurate carding and judgment of industry scale and customer needs. "Technical toB is a long-term and determined business." Chen Dingjia, chief technology officer of Kuaishou Technology, said at the launch of B-end business brand StreamLake on Aug. 10.

"the really good technology is to go out."The enterprise service market has always been a difficult bone for domestic entrepreneurs. From the perspective of the model, most of the toB business has large investment, high cost and long cycle, and even the top domestic cloud manufacturers need more than a decade of long accumulation to barely make a profit. For entrepreneurs, even if they borrow good business models from abroad, they will not be satisfied in China because of user education, willingness to pay, enterprise life cycle and other problems. In short, toB is a "bitter" and "heavy" business. In early August, Yu Bing, senior vice president of Kuaishou Technology and head of StreamLake, at Kuaishou Technology's headquarters in Xier Banner, Beijing, accepted an exclusive interview with 36Kr Holdings.

Yu Bing, senior vice president of Kuaishou Technology and head of StreamLake, "Why did Kuaishou Technology enter toB at this point in time?" After a brief greeting, 36Kr Holdings threw the issue that everyone was most concerned about to him. Yu Bing is confident. He joined Kuaishou Technology for six years, leading the audio and video business, which is the base of Kuaishou Technology's technology, and had previously been involved in two start-ups as a technology partner-a typical executive who had "heard the sound of gunfire on the front line". Yu Bing uses the three dimensions of "want to do, can do, can do" to answer the reason why Kuaishou Technology joined the toB and launched StreamLake. Want to do, from Kuaishou Technology's "audio and video + AI team" confidence in their own technology, in addition to the high-density talent structure, what is more important is years of technology accumulation and patent reserves, which makes Kuaishou Technology in almost all competitions such as peer competition in an invincible position. To be able to do so is to get positive feedback from the business model after technical verification. Yu Bing talked about "ROI" to 36Kr Holdings many times. As one of the leading video platforms in China, Kuaishou Technology's own bandwidth is close to hundreds of TB and has a huge investment. It has reduced a considerable proportion of the cost through a series of technology and algorithm optimization, and this obvious positive ROI result also makes enterprise customers more convinced of StreamLake's program. What can be done is Kuaishou Technology's judgment on the market industry. According to iResearch, the overall market size of video cloud services in China will reach 44.8 billion yuan in 2021 and is expected to grow at a compound annual growth rate of 29.5% in the next three years. For Kuaishou Technology, to cut into the video cloud market through audio / video + AI, even if you get a 10% share, it is still billions of dollars of space.

"the cake in this space is getting bigger and bigger, depending on whether the knife is fast or not. I think our knife is quite fast." Yu Bing told 36Kr Holdings that StreamLake opened up the core competencies such as infrastructure, AI, audio and video, and algorithms accumulated by Kuaishou Technology to other enterprises through modularization and standardization to form full-link services. From the point of view of customers, domestic enterprises have a particularly strong demand for "video" upgrading, and media upgrading pain points such as Zhihu Inc. are common. Mutual entertainment, social interaction, education, e-commerce, real estate, cars. Whether it is the Internet segment industry or traditional areas, are accelerating from graphics and text to video. Video upgrade needs to seize the window, rapid deployment online is the premise, and then from have to excellent. The problem is that video is a well-deserved "heavy industry". Generally speaking, a company needs to prepare for its video business for half a year to a year. It is normal to spend tens of millions or even hundreds of millions of dollars, and it is necessary to continue to invest. More importantly, talents in the video industry are very scarce. It takes a high time and labor cost to set up a team. StreamLake can achieve full link optimization experience and cost from codec standards, players, algorithms to intelligent video processing chips. "We just want to upgrade video from 'heavy industry' to 'light industry'." Yu Bing believes that the core value of StreamLake lies in two points: improving experience and controlling costs. Kuaishou Technology standardizes and encapsulates the solution into an operating system that allows customers to access easily and solve the problems of high bandwidth cost, innovation and low scale efficiency. From wanting to do, being able to do, to being able to do, from internal technical carding to external customer verification, Kuaishou Technology is fully prepared to enter toB. Before the official launch of StreamLake, Kuaishou Technology CTO Chen Dingjia said to the team: "the really good technology is not to say that you are good at home, but to go out to Huashan to discuss the sword, to compete with leading companies in the industry, and customers are willing to pay."

Reshape toB logicIn June 2022, Kuaishou Technology executives had a separate discussion around the toB business, "very much like MBA in-depth analysis of a case." Yu Bing recalls. A core proposition of this discussion is to explore and summarize the results of the StreamLake business model in stages, as well as greater possibilities. How did traditional cloud vendors develop? There is already a set of clear templates in the industry, starting from underlying computing, storage and other basic cloud resources (IaaS), competing with each other through spending money and price war, winning the scale, and then selling PaaS, SaaS and other software products on the upper floor; and because it is difficult for software products to scale up, eventually cloud manufacturers have to go back to the starting point and sell back traditional computing resources. The result is a sustained high loss, even if you can barely make a profit, it will take years of loss in exchange for it. Kuaishou Technology hopes to reshape a new set of business logic in the toB field. "after StreamLake's business model is injected into the industry, there will be innovations in the industry from loss to profit, which will bring a lot of changes to the industry." Yu Bing told 36Kr Holdings. First of all, when it comes to the billing model of cloud services, StreamLake is very different (or even diametrically opposed) from traditional cloud vendors, which Yu Bing describes as "co-directional logic with customers". In the case of Zhihu Inc., if it is the solution of a traditional cloud vendor, you only need to increase the bit rate for the purpose of improving clarity, but this means that the bandwidth is soaring and the cost is very high. However, the core of the technical service provided by StreamLake for Zhihu Inc. is to make the video on Zhihu Inc. platform both small (smooth and does not occupy large bandwidth) and clear through high-quality algorithms and compression technology, and finally achieve controllable bandwidth cost on the premise of ensuring picture quality. Reflected in the business model, StreamLake's plan not only charges part of the technical fees, but also adopts a win-win business model for the customer-side bandwidth fees that can be saved, which means that StreamLake wants to achieve more revenue, which means that it needs to help customers save a lot of bandwidth costs, which is really with the customer's business goals and become a "life community". "Border sense" is also another feature of Kuaishou Technology's toB business. Cloud manufacturers in the industry generally pursue "large and comprehensive" solutions. Architecturally, from the bottom IaaS to the upper SaaS; industry, traditional cloud manufacturers can be seen from the Internet, finance, and industry. This is naturally driven by the giant's expectation of revenue (or even profit) of the cloud business after the frenzied investment of money in the early stage. With a variety of lessons from the forerunners, Kuaishou Technology also has a clear positioning for the "do or not to do" of the toB business. "Let's just focus on one problem-- video." Yu Bing stressed to 36Kr Holdings. In addition to the focus on products, in the business model, Kuaishou Technology also pays attention to avoid blind expansion, burning money, but adhere to healthy, positive, and even profitable B-end business. According to 36Kr Holdings, StreamLake already has an independent technical production, research, sales and implementation team; at the beginning of 2022, Kuaishou Technology StreamLake has achieved positive revenue, which makes Yu Bing and his team more confident about their business strategy and direction. Of course, as a young "start-up" team incubated within the company, StreamLake still has a lot of "make-up lessons" work, such as how to improve the team's toB methodology in a toC-based business atmosphere, and how to better understand and digest customer needs, all of which make Yu Bing "feel like going back to the market." From toC to toB, new thinking is slowly igniting StreamLake. Kuaishou Technology headquarters, a StreamLake team member's desk on the "give customers a reason-Huawei sales negotiation and communication skills", in the past such a book almost impossible to appear on the desktop of Kuaishou Technology technical staff, "toB mysterious and daily, the real master is in every link and detail to the extreme." The StreamLake employee wrote on his sharing notes. As for the future of StreamLake, Yu Bing talked about Snowflake--, a unicorn born in Silicon Valley. Snowflake Inc achieved great commercial success after doing a good job of doing a good job of cloud native database. Snowflake Inc's share price soared 104% at the start of trading in September 2020 and now has a market capitalization of more than $50 billion. "We want to be Snowflake Inc in the video field." Yu Bing told 36Kr Holdings, "We talk to our customers about audio and video + AI, and we solve these two problems. We don't do anything else. As long as we upgrade the video, do the gathering and distribution, and find StreamLake." Simple, but extremely, this is Yu Bing's expectation of StreamLake, in fact, he has already hidden this label in the name of StreamLake. If translated literally, StreamLake means "stream lake". Yu Bing told 36Kr Holdings that what Kuaishou Technology has done since its establishment is to gather huge amounts of video content and distribute it to hundreds of millions of users. From content to users, the flow is like a trickle into the lake. "this is what we have been doing for more than a decade." He said. The following is a transcript of the conversation between 36Kr Holdings and Yu Bing, which has been deleted by the editor and released:36Kr Holdings: why did you choose to launch StreamLake at this node?Yu Bing:Since 2016, Kuaishou Technology has accumulated a number of leading experts in the industry. up to now, there are hundreds of people. at first, many technologies were connected to third parties by Kuaishou Technology, and later, due to the requirements of experience, they gradually turned to self-research. Then experienced the explosion of Kuaishou Technology's business, and now it has reached a more mature stage. From the point of view of serving customers, StreamLake is divided into "three steps". He initially served Kuaishou Technology's internal "customers", some independent APP and products, and later tried to support some external customers, such as the "Rain classroom" of Tsinghua University, and then to a large-scale platform like Zhihu Inc.. Dingjia often tells us that the really good technology is not to say that you are good at home, but to go out to Huashan to discuss the sword and really compete with the leading companies in the industry. Customers are willing to pay, and the ones that can make money are good technologies.36Kr Holdings: what signs or phenomena will make you feel that it is mature to launch toB business now?Yu Bing:Over the past two years, StreamLake has served a large number of customers, validating technical capabilities, product capabilities and business models. Kuaishou Technology started late in the B-end business, but he thought about it very clearly. There are a large number of traditional enterprises and platform enterprises in China, which have a strong demand for digital upgrading and video transformation. This is a very large incremental market and an appropriate time to enter. The core thing is that our technology is doing very well, and we are not at a disadvantage in all kinds of competitions.36Kr Holdings: for the interior of Kuaishou Technology, how do you evaluate the ROI that entered toB? What kind of internal discussions have you experienced?Yu Bing:Kuaishou Technology is very optimistic about the potential and opportunities in the field of toB. But toB itself needs a long period, unlike the C end, which can become popular in a few months. ToB needs hard work and a harvest. We have long-term patience, which is different from some of our peers. In the process of laying out new business, we will fully respect the law of toB, polish the business together with customers, and constantly expand the scale of customers through word-of-mouth communication. We do not pursue speed, nor do we burn money for growth.36Kr Holdings: let's talk about the start-up phase of the project. Zhihu Inc. is your early customer. How did you start to cooperate?Yu Bing:In 2020, Zhihu Inc. will accelerate media upgrading and strengthen the integration of picture and text and video. In the process of seeking external partners, based on the understanding and recognition of Kuaishou audio and video technology, we have been found twice. In the second year, we began to work together around architectural upgrades and experience optimization. Kuaishou Technology helps Zhihu Inc. solve the needs of user experience enhancement and cloud service cost optimization through media processing services, integration of CDN, player and other products, combined with hot trigger and multi-rate adaptive strategies. Zhihu Inc. originally used the previous generation H.264 codec standard, which everyone knows that the compression ratio is the codec standard of every generation, from H.264 to H.265 to H.266, each generation will be 40% better than 50%, Zhihu Inc. wants to use H.265, but found that there are many obstacles. First of all, you need to adapt to many terminal models, but there are too many single models. The second is the solution of traditional cloud manufacturers. If you use H.265, the transcoding service will cost a lot of money, and there is no good solution to help them save this money. On the one hand, Kuaishou Technology developed the KVC codec standard, which saves 30% to 50% more bandwidth than H.265; on the other hand, we also have a hot trigger technology, not all videos are worth decompressing, we actually only deal with hot videos, and ROI is positive, and the computing power and bandwidth ratio spent on it are cost-effective. We have also made a machine learning model to predict whether this thing will be hot, and if it will be hot, it will be compressed in advance, and compression will take time. Each video has its own life cycle, and it will be compressed before the video popularity reaches its peak. You can save the bandwidth.36Kr Holdings: won't cloud manufacturers make such a plan?Yu Bing:It is not easy, one is the model adaptation, the other is the heat trigger technology, all need a large amount of real data to do training, we talk to Zhihu Inc. about this plan, they feel suddenly enlightened. After testing, our compression is 42% better than the H.265 provided by cloud vendors. There is a reason behind this. Cloud manufacturers do this thing ROI is negative, we do ROI is particularly positive, cloud manufacturers take bandwidth as the main revenue target, (the more compression), their revenue will be reduced. While the logic of our customers is in the same direction, toB should follow the logic of the same direction as far as possible with customers, rather than the reverse logic. If I price your business in the way of traditional cloud manufacturers, I am also in the opposite direction. The more I bet, the more I lose, the less my income. As a result, we designed a very novel cooperation model, so that we have more motivation to iterate the compression algorithm to bring business benefits.36Kr Holdings: is the plan provided by Kuaishou Technology to Zhihu Inc. customized or standardized?Yu Bing:Is a standardized product, but we will do more in the service, but also directed to help Zhihu Inc. do the first screen optimization, this is actually our differentiation advantage, we can do very professional business-based (service). It's the equivalent of consulting experts to help them diagnose problems. Zhihu Inc. 's pain points and problems are common in the market, it represents a series of media upgrading process of customer needs, we make a general solution, will better copy and support other customer transformation.36Kr Holdings: did you think there were many similar customers in the market at that time?Yu Bing:Quite a lot, we also set up, video is the theme of the Internet in the past and the next few years. Video not only includes short videos, but also live streaming and interaction, which will be landed in more industries. For example, live e-commerce, it is much higher than the traditional shelf e-commerce conversion rate, video + or live + is a very surging wave. In this context, we also talked about a lot of customers. Internet customers are considering adding video modules to their business, and are thinking about video live streaming and real-time interaction, how to improve the conversion rate of stock business and help them innovate. If you don't do this, you'll be more anxious.36Kr Holdings: but many giants' own video technology needs will be digested with their own cloud business.Yu Bing:The head video platform is not our key consideration. They are sure that a large number of technologies are used by themselves. For many customers in the middle waist, Zhihu Inc. is still very large, and there are hundreds or thousands of them in the middle waist, and their business volume is not small. The video itself belongs to heavy industry, with high input and high threshold. We turn it into several very simple interfaces, which can encapsulate a very heavy and complex system, but for customers, it becomes a matter of easy access to the interface.36Kr Holdings: after the initial verification, does StreamLake have a clear project node?Yu Bing:In June this year, Kuaishou Technology's internal management discussed, a bit like MBA in-depth analysis of the case, a core discussion proposition is: phased discussion and summary of B-side business results, as well as greater possibilities. Although cloud manufacturers have high bandwidth revenue, they all lose money, and none of them make money. Our business model is injected into the industry, which can make the industry change from loss to profit. We think this business is worth doing, and it is done as an independent brand.36Kr Holdings: what is Kuaishou Technology's position on StreamLake? Do you want you to make money right away, or are you allowed to explore for a while?Yu Bing:There is room for exploration. Kuaishou Technology still has its rules in doing things (especially toB business). Kuaishou Technology's investment in the early stage will not be very large, and we hope to gradually expand the scale after verification.36Kr Holdings: many giants will hold high in the early stage of toB business, Kuaishou Technology seems to be very cautious?Yu Bing:At this stage, the industry does not have this kind of understanding, before is to seek scale, work quickly, unhealthy, which will give birth to a lot of very low ROI or even loss-making business. The business is also lack of added value, buying and selling, or completely customized development, which does not have much value in the toB market. What is really valuable is to provide added value with high added value and continuous high efficiency, which is good. We want to be Snowflake Inc (American Enterprise Services listed company, focusing on database business) in the video field. We chat with customers and talk about video + AI. The core is to solve these two problems and do nothing else, so customers just want to find StreamLake for video transformation, aggregation and distribution.36Kr Holdings: the name of StreamLake is what you got?Yu Bing:When we named it, we thought of two words, one is called gathering and the other is called distribution. Kuaishou Technology has been doing the gathering and distribution of content from the very beginning. A huge amount of content is sent to it every day, and Kuaishou Technology distributes it to hundreds of millions of users. This is what we have been doing for more than a decade. One of its technical bases is called video, and the other is called AI,Video+AI, the two core technical bases for us to do this. So we start from this, thinking about the intention of Video and AI to converge and distribute, thinking that it is like a lake (Lake), with a trickling stream upstream and the Yangtze River and Yellow River downstream.