Going overseas is an unavoidable question for Chinese companies at different stages of development.
According to the latest China Council for the Promotion of International Trade (CCPIT) Research Institute report, China's foreign direct investment flow and stock have ranked among the top three in the world for four consecutive years. At the same time, nearly 80% of Chinese enterprises will maintain and expand their foreign investment intentions and are optimistic about the prospects for foreign investment. According to Accenture's "China Unicorn Research", 45% of unicorn companies believe that overseas expansion is crucial, and 82% of them plan to expand overseas within 24 months.
In terms of layout, China Bridge Consulting’s research report shows that more than 60% of the companies have gone overseas and have expanded their overseas routes from traditional Europe and the United States to South America, the Middle East, Africa and other regions. Southeast Asia, Europe and North America are still the preferred destinations for Chinese companies to go overseas. At the same time, emerging regions such as South America, Africa, and the Middle East are becoming the "new blue ocean" for Chinese companies to go overseas.
"From a macro perspective, companies going overseas will face the impact of different market policies and regulations; in terms of growth, companies need to customize their overseas products and even their promotion paths based on a deep understanding of cultures and audiences in different regions. Which one to choose? Media distribution channels? How to customize appropriate marketing materials according to the audience's preferences? What tools are used to improve the ROI of overall marketing? These are the challenges that companies will face." said in the interview.
Mobvista is a global technology platform. One of its main businesses is to provide advertising services for global app developers, usually through programmatic advertising transactions. According to the latest "Comprehensive Performance Report of Global Advertising Platforms" released by AppsFlyer, a global marketing measurement and experience management platform, Mobvista's Mintegral programmatic advertising platform ranks fourth in the IOS and Android dual-end retention strength list of all categories of applications, and is one of the top five in the world The only Chinese platform in .
After Apple's new privacy regulations, how will its performance grow against the trend?
The programmatic advertising mentioned here can be understood as using AI to accurately match people, scenes and advertisements to improve the matching efficiency between users and advertisers. However, with the enhancement of user privacy awareness, the difficulty of obtaining user data continues to increase. A landmark event is the Apple IDFA New Deal in 2021.
IDFA (Identifier for Advertising) is an advertising identifier, which is the only official Apple official permission to allow advertisements to obtain user information. For example, when a user searches for coffee in a shopping app, the IDFA records the search content. When the user opens the news app again, the news app will upload the user's IDFA, and the shopping app sees this IDFA and puts coffee on the news app. advertise. After that, the shopping app will also use IDFA to track whether users have actual purchase behavior after receiving advertisements.
After Apple's new IDFA policy, applications can only use IDFA if the user actively clicks "authorization and consent". But in fact, even if IDFA is desensitized, most users still believe that refusing to track is equivalent to protecting their privacy. According to the statistics of Single, a member of the Post-IDFA Alliance, only 16.8% of users have chosen to accept ad tracking. If the platform cannot obtain IDFA, it will be a fatal blow to the mobile digital advertising ecosystem within Apple's system.
Google, Meta, etc. rely on IDFA to obtain user portraits, while the algorithms of third-party advertising platforms such as Mintegral mainly model context and do not rely on user data. According to AppsFlyer's report, after experiencing great changes in the mobile advertising ecosystem, the situation of "two powers coexisting" dominated by Meta (ie Facebook) and Google in the past is changing to "coexistence of giants".
"After the release of Apple's new privacy regulations, the industry has undergone a major reshuffle. AppsFlyer's "Advertising Platform Comprehensive Performance Report" (14th edition), which is more concerned in the industry, shows that in the mobile advertising platform retention index list and IAA index list, The rankings of third-party advertising platforms, including Mintegral, are higher than Meta and Google. From this perspective, the budget allocation of advertisers will change in the future, and more advertising budgets will flow out of the head media platform and transfer to independent third-party programmatic platform." Xi Yuan told The Paper (www.thepaper.cn).
Data shows that in 2021, when the advertising revenue of leading Internet companies is slowing down, Mintegral, Mobvista's core programmatic advertising platform business, achieved revenue of US$575 million, a year-on-year increase of 88.0%.
When answering how to achieve performance growth against the trend, Xi Yuan told The Paper (www.thepaper.cn), "Internally, platform development needs to take a long-term route, and it is inseparable from algorithms and creative resonance. The modeling system adopted is different from the current general IDFA algorithm in the entire industry. It does not rely on traditional user privacy data, but is based on short-term context instant interest modeling. At the same time, Mintegral is also investing in dynamic creative optimization, so that it does not rely on In the case of user privacy data, it can still achieve better conversion results.”
At the same time, in terms of algorithm modeling, Mintegral is more about building a user cluster, that is, not relying on the data of a single user, but the common characteristics of a group of users to improve the accuracy of advertising.
According to Xi Yuan, Mobvista started to carry out a number of compliance certifications for its business segments in 2017. From the perspective of short-term benefits, spending a lot of time and cost on data security and privacy protection that is difficult to see in the short term seems to be a thankless sign. But in fact, Xi Yuan believes that the one step ahead in compliance and security has provided important support for the rapid expansion of the global market and the construction of the SaaS tool ecosystem.
"Mintegral has always invested a lot of resources on the algorithm side. We have deployed a huge algorithm cluster in the cloud and continue to recruit top talents in the industry. In 2021, Zhu Xiaoqiang, the algorithm leader of Alimama's display advertising, will join Mobvista as the chief executive officer. The artificial intelligence officer also provides stronger support for Mintegral's algorithm system iteration. In fact, third-party advertising platforms rarely make such a high investment in algorithms, because in the eyes of many companies, the advertising industry is sales, operation or Creativity-driven." Xi Yuan said.
Cloud native: how to operate in more than 200 countries and regions around the world?
Mobvista's network alliance, programmatic advertising platform and other businesses need to provide 24-hour uninterrupted mobile marketing solutions for more than 200 countries and regions around the world. Over 100 billion ad requests are processed daily through algorithmic predictions. Behind this is the data processing scale of tens of petabytes per day, and the model training scale of billions of features per hour. These data have strong sparsity and complex data dimensions. According to reports, Mobvista's DNN (Deep Neural Network) model has a huge sparse embedding layer, containing features with more than 10 billion dimensions.
At the same time, the advertising industry has extremely high demand for data timeliness. The complex algorithm behind each ad request predicts that the end-to-end data latency requirement must be less than 50 milliseconds.
How is data processing at this scale achieved?
Behind this is the cloud service industry with a rapidly growing market size.
Mobvista is known as the "first stock in mobile marketing". When it was founded in 2013, it chose cloud native. At that time, "cloud" was still a rare concept. When answering why he made such a choice at the time, Xi Yuan told The Paper (www.thepaper.cn), “Mobvis’ business has experienced a period of rapid growth, and the average daily ad requests surged from 18 billion in 2017 to 2018. About 50 billion times a year, and it will continue to grow to 60 billion times in 2019, reaching a maximum of 100 billion times. At the same time, we are conducting business overseas, and to achieve low response delay services, it is impossible for us to build our own in every country. Data center. The globalized services of Amazon Cloud Technology can help us achieve dynamic matching of resources and business scale while the business scale is growing rapidly.”
Through cloud service providers, Mobvista mainly solves two problems.
First, build a one-stop big data machine learning platform system.
Second, the mobile advertising business is spread all over the world, so how to respond to user requests from all over the world with the fastest speed and implement the global deployment of the system is a very big challenge. Moreover, such services have extremely high requirements on the stability of the system, because traffic peaks often occur, and the instantaneous click volume will rise sharply, which requires the system to be able to process quickly and have very good scalability.
With the growth of business, the cost of using cloud will become the biggest trouble. The development of cloud computing has put forward higher requirements for software architecture. The architecture needs to have the functions of fast response and elastic scaling, on-demand acquisition, resource maintenance and hosting, and realize cost control. With the growth of business requirements, the code structure tends to be complex, and the time cost from forming the code to completing the deployment process increases, that is, "bigger is slower"; the traffic brought by business development increases, which makes the enterprise cost more on the machine load. The increase in demand increases, that is, "bigger is more expensive".
"We have always been optimizing the cost of the cloud. Since 2017, we have set up a special team to do research in this area." Xi Yuan specifically shared two ways to reduce costs:
On the one hand, Amazon Cloud Technology's cloud resources, especially computing resources, are divided into different categories. "We are now using some Amazon EC2 Spot Instance (computing instance) resources. For example, if you don't know the usage of next year or the year after, there are two solutions. One is to purchase pre-orders in advance. Reserve an instance as a backup. However, because you may not be able to predict your own calculation amount, you may waste resources, so we buy more according to demand. At the earliest time, we bought a lot of OD (On Demand, press There is also a Spot Instance, its biggest feature is that the price can be as low as 10% of the OD, so we must make good use of this resource and find ways to use it to develop innovative products.”
In addition, it is to optimize the use of computing power. "For example, if you have a room in your house that is empty, you will definitely turn off the lights. For us, we have to analyze the tens of billions of ad requests online every day, which of them are necessary to calculate Yes. Some requests may be analyzed through historical analysis, and you will find that every time you go to bid, you will not win the bid, and the entire work process will actually be wasted. Therefore, we will sort it out. At present, some ad requests may not be bid. This is equivalent to putting Internal work is optimized, not just using cheap resources.
According to the latest China Council for the Promotion of International Trade (CCPIT) Research Institute report, China's foreign direct investment flow and stock have ranked among the top three in the world for four consecutive years. At the same time, nearly 80% of Chinese enterprises will maintain and expand their foreign investment intentions and are optimistic about the prospects for foreign investment. According to Accenture's "China Unicorn Research", 45% of unicorn companies believe that overseas expansion is crucial, and 82% of them plan to expand overseas within 24 months.
In terms of layout, China Bridge Consulting’s research report shows that more than 60% of the companies have gone overseas and have expanded their overseas routes from traditional Europe and the United States to South America, the Middle East, Africa and other regions. Southeast Asia, Europe and North America are still the preferred destinations for Chinese companies to go overseas. At the same time, emerging regions such as South America, Africa, and the Middle East are becoming the "new blue ocean" for Chinese companies to go overseas.
"From a macro perspective, companies going overseas will face the impact of different market policies and regulations; in terms of growth, companies need to customize their overseas products and even their promotion paths based on a deep understanding of cultures and audiences in different regions. Which one to choose? Media distribution channels? How to customize appropriate marketing materials according to the audience's preferences? What tools are used to improve the ROI of overall marketing? These are the challenges that companies will face." said in the interview.
Mobvista is a global technology platform. One of its main businesses is to provide advertising services for global app developers, usually through programmatic advertising transactions. According to the latest "Comprehensive Performance Report of Global Advertising Platforms" released by AppsFlyer, a global marketing measurement and experience management platform, Mobvista's Mintegral programmatic advertising platform ranks fourth in the IOS and Android dual-end retention strength list of all categories of applications, and is one of the top five in the world The only Chinese platform in .
After Apple's new privacy regulations, how will its performance grow against the trend?
The programmatic advertising mentioned here can be understood as using AI to accurately match people, scenes and advertisements to improve the matching efficiency between users and advertisers. However, with the enhancement of user privacy awareness, the difficulty of obtaining user data continues to increase. A landmark event is the Apple IDFA New Deal in 2021.
IDFA (Identifier for Advertising) is an advertising identifier, which is the only official Apple official permission to allow advertisements to obtain user information. For example, when a user searches for coffee in a shopping app, the IDFA records the search content. When the user opens the news app again, the news app will upload the user's IDFA, and the shopping app sees this IDFA and puts coffee on the news app. advertise. After that, the shopping app will also use IDFA to track whether users have actual purchase behavior after receiving advertisements.
After Apple's new IDFA policy, applications can only use IDFA if the user actively clicks "authorization and consent". But in fact, even if IDFA is desensitized, most users still believe that refusing to track is equivalent to protecting their privacy. According to the statistics of Single, a member of the Post-IDFA Alliance, only 16.8% of users have chosen to accept ad tracking. If the platform cannot obtain IDFA, it will be a fatal blow to the mobile digital advertising ecosystem within Apple's system.
Google, Meta, etc. rely on IDFA to obtain user portraits, while the algorithms of third-party advertising platforms such as Mintegral mainly model context and do not rely on user data. According to AppsFlyer's report, after experiencing great changes in the mobile advertising ecosystem, the situation of "two powers coexisting" dominated by Meta (ie Facebook) and Google in the past is changing to "coexistence of giants".
"After the release of Apple's new privacy regulations, the industry has undergone a major reshuffle. AppsFlyer's "Advertising Platform Comprehensive Performance Report" (14th edition), which is more concerned in the industry, shows that in the mobile advertising platform retention index list and IAA index list, The rankings of third-party advertising platforms, including Mintegral, are higher than Meta and Google. From this perspective, the budget allocation of advertisers will change in the future, and more advertising budgets will flow out of the head media platform and transfer to independent third-party programmatic platform." Xi Yuan told The Paper (www.thepaper.cn).
Data shows that in 2021, when the advertising revenue of leading Internet companies is slowing down, Mintegral, Mobvista's core programmatic advertising platform business, achieved revenue of US$575 million, a year-on-year increase of 88.0%.
When answering how to achieve performance growth against the trend, Xi Yuan told The Paper (www.thepaper.cn), "Internally, platform development needs to take a long-term route, and it is inseparable from algorithms and creative resonance. The modeling system adopted is different from the current general IDFA algorithm in the entire industry. It does not rely on traditional user privacy data, but is based on short-term context instant interest modeling. At the same time, Mintegral is also investing in dynamic creative optimization, so that it does not rely on In the case of user privacy data, it can still achieve better conversion results.”
At the same time, in terms of algorithm modeling, Mintegral is more about building a user cluster, that is, not relying on the data of a single user, but the common characteristics of a group of users to improve the accuracy of advertising.
According to Xi Yuan, Mobvista started to carry out a number of compliance certifications for its business segments in 2017. From the perspective of short-term benefits, spending a lot of time and cost on data security and privacy protection that is difficult to see in the short term seems to be a thankless sign. But in fact, Xi Yuan believes that the one step ahead in compliance and security has provided important support for the rapid expansion of the global market and the construction of the SaaS tool ecosystem.
"Mintegral has always invested a lot of resources on the algorithm side. We have deployed a huge algorithm cluster in the cloud and continue to recruit top talents in the industry. In 2021, Zhu Xiaoqiang, the algorithm leader of Alimama's display advertising, will join Mobvista as the chief executive officer. The artificial intelligence officer also provides stronger support for Mintegral's algorithm system iteration. In fact, third-party advertising platforms rarely make such a high investment in algorithms, because in the eyes of many companies, the advertising industry is sales, operation or Creativity-driven." Xi Yuan said.
Cloud native: how to operate in more than 200 countries and regions around the world?
Mobvista's network alliance, programmatic advertising platform and other businesses need to provide 24-hour uninterrupted mobile marketing solutions for more than 200 countries and regions around the world. Over 100 billion ad requests are processed daily through algorithmic predictions. Behind this is the data processing scale of tens of petabytes per day, and the model training scale of billions of features per hour. These data have strong sparsity and complex data dimensions. According to reports, Mobvista's DNN (Deep Neural Network) model has a huge sparse embedding layer, containing features with more than 10 billion dimensions.
At the same time, the advertising industry has extremely high demand for data timeliness. The complex algorithm behind each ad request predicts that the end-to-end data latency requirement must be less than 50 milliseconds.
How is data processing at this scale achieved?
Behind this is the cloud service industry with a rapidly growing market size.
Mobvista is known as the "first stock in mobile marketing". When it was founded in 2013, it chose cloud native. At that time, "cloud" was still a rare concept. When answering why he made such a choice at the time, Xi Yuan told The Paper (www.thepaper.cn), “Mobvis’ business has experienced a period of rapid growth, and the average daily ad requests surged from 18 billion in 2017 to 2018. About 50 billion times a year, and it will continue to grow to 60 billion times in 2019, reaching a maximum of 100 billion times. At the same time, we are conducting business overseas, and to achieve low response delay services, it is impossible for us to build our own in every country. Data center. The globalized services of Amazon Cloud Technology can help us achieve dynamic matching of resources and business scale while the business scale is growing rapidly.”
Through cloud service providers, Mobvista mainly solves two problems.
First, build a one-stop big data machine learning platform system.
Second, the mobile advertising business is spread all over the world, so how to respond to user requests from all over the world with the fastest speed and implement the global deployment of the system is a very big challenge. Moreover, such services have extremely high requirements on the stability of the system, because traffic peaks often occur, and the instantaneous click volume will rise sharply, which requires the system to be able to process quickly and have very good scalability.
With the growth of business, the cost of using cloud will become the biggest trouble. The development of cloud computing has put forward higher requirements for software architecture. The architecture needs to have the functions of fast response and elastic scaling, on-demand acquisition, resource maintenance and hosting, and realize cost control. With the growth of business requirements, the code structure tends to be complex, and the time cost from forming the code to completing the deployment process increases, that is, "bigger is slower"; the traffic brought by business development increases, which makes the enterprise cost more on the machine load. The increase in demand increases, that is, "bigger is more expensive".
"We have always been optimizing the cost of the cloud. Since 2017, we have set up a special team to do research in this area." Xi Yuan specifically shared two ways to reduce costs:
On the one hand, Amazon Cloud Technology's cloud resources, especially computing resources, are divided into different categories. "We are now using some Amazon EC2 Spot Instance (computing instance) resources. For example, if you don't know the usage of next year or the year after, there are two solutions. One is to purchase pre-orders in advance. Reserve an instance as a backup. However, because you may not be able to predict your own calculation amount, you may waste resources, so we buy more according to demand. At the earliest time, we bought a lot of OD (On Demand, press There is also a Spot Instance, its biggest feature is that the price can be as low as 10% of the OD, so we must make good use of this resource and find ways to use it to develop innovative products.”
In addition, it is to optimize the use of computing power. "For example, if you have a room in your house that is empty, you will definitely turn off the lights. For us, we have to analyze the tens of billions of ad requests online every day, which of them are necessary to calculate Yes. Some requests may be analyzed through historical analysis, and you will find that every time you go to bid, you will not win the bid, and the entire work process will actually be wasted. Therefore, we will sort it out. At present, some ad requests may not be bid. This is equivalent to putting Internal work is optimized, not just using cheap resources.
Comments