vivo platform practice exploration journey-platform product series 01

Author | Yang Yang, vivo Internet Platform Product R&D Team

This is the first in the series of articles on vivo platform products. It mainly describes how vivo software engineers gradually solved thorny problems such as software reuse and rapid response to business changes through system platform construction and other means in the context of rapid business development. We have selected representative platform products such as content review platform, AB experiment platform, version release platform, etc., and provide a detailed introduction to the platformization experience.

1. Platformization

1.1 Business Status

Vivo's Internet business started in 2011. After years of development, the daily active users of Vivo's core Internet applications have reached more than 10 million, and the cumulative number of Vivo users has reached 500 million by 2022. Represented by application stores, browsers, videos, game centers, wallets, and official website malls, an Internet business model with advertising, distribution, and games as the core has gradually been formed.

1.2 Problem

Today, large Internet companies have a wide range of business types, with multiple business units developing in parallel and the business becoming increasingly complex. Companies are also constantly upgrading their technology and adjusting their business, such as introducing microservice architecture and redivision of business boundaries. The evolution of microservices makes it difficult to reuse public capabilities coupled to business systems. The redivision of business boundaries is often accompanied by further refinement of the organizational structure, which can lead to repeated solutions to the same type of problems and waste of resources. Therefore, how to extract the common capabilities of multiple business lines, centrally manage and evolve them, avoid repeated investment, and quickly support business iteration has become an urgent problem to be solved within the company.

1.3 Platformization

The evolution of platform-based architecture is an inevitable choice when the business develops to a certain scale. Its core is to decouple problems in a certain field from the business software and aggregate them into an independent platform for centralized management and control, and continuously evolve a platform ecosystem with gradually clear standards and boundaries, so as to solve a class of problems in a one-stop manner, achieve efficient reuse within the organization, and support rapid business iteration.

In the past few years, vivo Internet has experienced systematic technology construction, from the single chimney of each team to service-oriented, and now to platform-oriented, vivo has gradually formed its own platform ecosystem (as shown in Figure 1). In the process, we have also encountered many challenges in technology, organization and other aspects. We are about to bring readers a series of special articles - vivo platform construction, and this article is the first article in the special series.

2. Excellent Case Studies

After years of development, vivo Internet has achieved remarkable results in platform construction, and has completed the construction of multiple platform products, covering multiple business areas. In terms of activity operation, there is the Wukong platform; in terms of quality control, there are content review platforms and monitoring platforms; in terms of strategy analysis, there are version release platforms and AB experiment platforms. This section mainly selects some excellent platform construction cases for a brief introduction, and the following series of special articles will provide a more comprehensive and in-depth introduction to these platforms.

2.1 Content Review Platform

2.1.1 Background

Before 2016, when vivo's Internet business was just beginning, each system that interacted with users (such as accounts, game forums, and official website communities) had established its own sensitive word filtering capabilities to implement simple illegal content blocking functions. With the rapid rise of vivo's Internet business, the content of the vivo ecosystem began to diversify, from the initial single text content to the current text, pictures, information, and videos. The content has undergone earth-shaking changes. The amount of text/information data is large, the illegal scenarios are complex, and it is extremely difficult for non-professional teams to review; the picture scenes are diverse, and illegal pictures emerge in an endless stream, especially those involving politics and pornography; the audio and video/live broadcast scenes are becoming increasingly rich, the quality of the anchors is uneven, and illegal content is rampant, which seriously affects the brand image.

2.1.2 Self-developed content review platform

Vivo's content was originally mainly reviewed by third-party platforms. With the rapid development of its business, this model can no longer meet the needs of Vivo's business development. In addition, the entire industry is increasing its investment in content review. In order to better ensure the security of Vivo's content, provide our users with a more extreme experience, and improve the reuse of system capabilities, Vivo decided to develop its own content review platform from scratch.

2.1.3 Current Status of Content Review Platform

The vivo content review platform has the capabilities of content security review, content quality identification, data annotation and public opinion collection. It uses the dual guarantee of "machine review + human review" to safeguard the security and quality of the business. The machine review capabilities mainly include AI judgment, black/white list system, sensitive word system, live streaming real-time monitoring mechanism and front-end reporting mechanism. The current accuracy rate of image and text judgment has reached 98%, which has reached the leading level in the industry. Manual review is equipped with hundreds of full-time personnel with corresponding professional knowledge and skills, who review and filter Internet information 24 hours a day, 7 days a week to ensure network information security.

2.2 AB Experiment Platform

2.2.1 Background

After experiencing a period of wild growth, Internet companies have gradually paid more attention to the scientific and refined development of products, transforming from extensive to intensive. In the United States, data-driven growth methodologies such as growth hacking are helping global technology giants such as Google, Microsoft, and Facebook achieve sustained business growth; in China, driving effective business growth through fine data operations and AB experimental analysis has gradually become a consensus and has become one of the core means of operating corporate stock data. Among them, AB experiments, as a typical representative, have naturally become an indispensable core operating means for mainstream domestic companies, effectively improving the conversion efficiency of traffic and the iteration efficiency of production and research.

2.2.2 What is AB experiment?

The so-called AB experiment is also called a "controlled experiment" or a "small-traffic random experiment". In order to verify the effect of a new strategy, two solutions are prepared: the original strategy A and the new strategy B. Then a small part of the total users is taken out and these users are completely randomly divided into two groups, so that there is no difference between the two groups of users from a statistical perspective. The original strategy A and the new strategy B are presented to different user groups respectively. After a period of time, the data is analyzed by combining statistical methods to obtain the change results of the indicators after the two strategies take effect, and to judge whether the new strategy B meets expectations.

2.2.3 vivo AB experimental platform

vivo Internet continues to attach importance to scientific experimental decision-making, which means that all changes to users must be based on the corresponding experimental conclusions. For example, modifying the background color of the top ad, testing a new ad click-through rate (CTR) prediction algorithm, and ultimately which solution is selected all require scientific decisions based on experimental results, so a powerful AB experiment platform becomes more and more important. In the past few years, vivo Internet has gradually embarked on the path of platform construction from a small experimental system with many chimneys - building and improving the AB experiment platform. The AB experiment platform has grown from a single system to a company-level one-stop platform for solving A/B experiment-related problems, helping the Internet's core business to conduct fast and accurate experiments and efficiently promote business growth.

2.3 Version Release Platform

2.3.1 What is a version release?

The version release platform, as the name implies, is mainly used to release mobile APK versions. Since the company developed mobile applications, system upgrades have been too heavy for users and the upgrade rate has been low. The rapid iteration momentum brought about by the rise of the Internet has made self-upgrade capabilities one of the indispensable capabilities. The vivo version release platform was originally built based on the self-upgrade of APK in mobile phones. It is no exaggeration to say that the vivo version release platform exists with vivo smartphones. The main business of the version release platform can be simply understood as: version transmission, rule configuration, and distribution package.

2.3.2 Business Scope

Of course, as the mobile phone business gradually expands and enriches, the scope of our platform, the granularity of its control, and the data analysis capabilities of its platform are becoming more and more perfect.

On the one hand, we have taken over many upgrade capabilities other than "APK self-upgrade", such as various components, quick applications, atomic components, SDK packages, plug-ins, SO libraries, etc.; on the other hand, we have made more compatible and detailed extensions to the distribution rules, such as quantity control and quantity stop, multi-dimensional positive and negative intersection and difference & grouping, identifier grouping, various grayscale strategies, multi-language, multi-channel, global rules, one-click interception, one-click analysis, etc.; in addition, security and compliance desensitization, cost reduction and incremental algorithms, flow control algorithms [Yugong Platform], Nebula burial point quality analysis, crash alarm analysis, release trends, etc. are also within the added value of the vivo version release platform.

2.3.3 Business objectives

Our goal is to build a full-link release platform for companies that is safe, compliant, intelligent, stable, efficient, and innovative. We want to provide a first-class user experience and set a benchmark for the reputation of company-level platform products. In this series of articles, we will focus on some of the explorations and practices of the vivo version release platform in the field of intelligent bandwidth optimization. We welcome readers to exchange and discuss.

2.4 Cloud Services

2.4.1 What is cloud service?

The vivo cloud service mainly provides two key capabilities. First, for C-end users, the vivo cloud service provides vivo mobile phone users with functions such as file backup and data synchronization. This part of the capability enables users to easily manage important data such as photos, videos, contacts, text messages, notes, etc. locally and in the cloud. In addition, the construction capabilities of the vivo cloud service also include the ability to provide fast and convenient C-end file upload and download capabilities for other businesses within vivo. At the same time, while building the data backup and synchronization capabilities for C-end users, we have built our own cloud service platform and improved the platform-related backup and synchronization functions. This platform function enables other brother businesses within vivo to quickly and conveniently use the capabilities of the cloud service platform to perform backup and synchronization capabilities for C-end users of their own businesses.

2.4.2 Cloud Services and Security

As we all know, cloud service businesses need to store massive amounts of data files for users. In addition to ensuring the stability of the service, special attention must be paid to data security, file persistence, etc. Many readers may also know that some manufacturers' cloud services have experienced user data leaks, and users' private files can be directly queried through search engines. This incident has a huge impact on corporate brands. In subsequent articles, we will introduce to readers in detail how vivo cloud service takes into account product functions, service stability and data security in software design.

2.4.3 Cloud Services and Cost Reduction and Efficiency Improvement

As the number of vivo mobile phone users continues to rise, the volume of files stored by users in the cloud is also growing. According to statistics, just three years after the cloud disk capability of vivo cloud service was completed, the total size of cloud file storage has exceeded 100 PB, and the growth rate can be said to be very rapid. Then the very severe test that follows is the cost issue. How can we control costs while ensuring that the user experience is not affected? This is a difficult and long-term practical problem, and we will bring you our solution later.

2.5 Points

2.5.1 What is a Point?

Points are an effective incentive tool for existing user operations on the Internet, carrying the mission of stimulating user activity and retention. As a points system, vivo points are also centered around acquisition and consumption, and the extended business capabilities include: points mall, gift center, check-in center, task center, and activity center.

2.5.2 Points Ecosystem

Vivo points are used throughout the entire vivo ecosystem for Internet applications. At the same time, the uniqueness of mobile phone manufacturers' Internet businesses (not limited to a single type of business) also makes vivo points different from other industry ecosystem points systems. These differences are mainly reflected in the close cooperation and mutual penetration of vivo points with various business forms. Usually, points are obtained based on tasks in the industry, and the completion of tasks is judged based on the completion of specific behaviors, and finally points are issued. Based on the "Fogg model", we can find three elements that motivate users to complete a certain behavior: motivation, conditions, and triggers. This provides theoretical support for the construction of tasks.

2.5.3 Points Tasks

As the most important way to obtain points, applications within the vivo ecosystem are connected to point tasks, guiding users to complete high-value behavioral tasks, which helps them achieve the core KPI indicators of the business side. At the same time, users are rewarded with points for completing tasks, which motivates them to continue to generate high-value behaviors, and ultimately forms a positive cycle. So how is vivo's task system built, and what process has the system construction gone through? In this series of articles, we will unveil its mysterious veil for everyone.

2.6 Account

2.6.1 What is an account?

For our users, the vivo account is a pass for users to enjoy the entire vivo ecosystem of services. The vivo ecosystem service content covers all aspects of life, games, entertainment, etc. At the same time, as an account system of a mobile phone manufacturer, it must shoulder the responsibility of protecting user device assets and ensuring the safety of users' mobile phones. For each business we serve, the account is the most basic platform service and the basic guarantee for the development of the business. And the vivo account not only serves our own mobile phone users, but also allows users to enjoy the convenient services brought by the vivo account in cross-platform and multi-ecological environments such as WeChat and iOS.

2.6.2 Account Globalization

With the expansion of the company's mobile phone business in overseas markets, accounts have been deployed across continents in seven major data centers around the world to serve our global users. The services of each data center strictly comply with the laws and regulations on personal data privacy and security in the service area. In order to cope with the risks brought by policy changes, the services in the GDPR area have the ability to isolate dynamic national data, etc., ensuring that certain service capabilities can be maintained under the interference of external environmental factors.

2.6.3 Stability Construction

With the rapid development of our business, the number of users we serve has exceeded 100 million very early, and the average daily call volume of our services has exceeded 10 billion. The account has become a typical system with three highs (high performance, high concurrency, and high availability). In addition to the positioning of basic platform services, a lot of work will be carried out around how to build the stability of the system. In this series of articles, we will talk with you about the attempts and efforts made by the account system in building stability.

3. Final Thoughts

This article is the first in a series of articles on vivo's platform construction. It aims to introduce the background of vivo's platform construction and the value of platform construction as well as some successful cases, in order to stimulate discussion. The following series of articles will introduce in depth the exploration, construction and innovation of vivo's Internet platform products, and share the problems encountered and thoughts in combination with the specific practices in the process. I hope to help readers understand the significance and methods of platform construction, feasibility assessment before construction, and traps that should be avoided in the process, etc.

After vivo carried out platform construction, it has achieved significant improvements in software reuse and R&D efficiency. However, platformization is not a silver bullet on the road of software architecture evolution that can cure all diseases. The platform may be difficult to meet the changes in business needs in the future due to insufficient consideration in the early stage or a relatively single business model in the early stage. Don't expect the platform designed at present to fully meet the unpredictable changes in the future. However, we cannot sit idly by for the uncertainty of the future. We need to make certain predictions about the development of future businesses and reserve extension points for the future. In addition, it is necessary to ensure that the extension points are decoupled from the current system as much as possible. The contradiction between platformization and business diversity will always exist, but at a certain stage this contradiction may be temporarily balanced.