小米

Baidao Helps Xiaomi with Data Analysis

Customer background

Xiaomi Technology Co., Ltd. was founded on March 3, 2010. It is a global mobile internet company and an innovative technology company that focuses on the research and development of smart hardware and electronic products, smartphones, smart electric vehicles, Internet TVs, and the construction of a smart home ecosystem. Xiaomi has created a model of developing mobile phone operating systems using the Internet model, with the participation of enthusiasts in development and improvement.

Xiaomi has built the world’s largest consumer IoT platform, connecting over 618 million smart devices and reaching more than 100 countries and regions around the world. The monthly active users of MIUI worldwide have reached 600 million. Xiaomi has invested in over 500 companies, covering areas such as smart hardware, consumer goods for daily life, education, games, social networks, culture and entertainment, healthcare, automotive transportation, finance, etc.

With the continuous expansion of its overseas business, its international business department has introduced Looker as a visualization platform for data analysis and business intelligence, based on the existing big data product ecosystem of Firebase, BigQuery, etc. on the Google Cloud Platform.

Business requirements and challenges

• 5 billion pieces of data are generated every day
• The total data volume is approximately 30 TB.
• The number of events included in a single business line is 400+
• Business indicator event attribute 300+
• Multiple business lines have data analysis requirements.
• Traditional data analysis methods usually require a large amount of time and human resources.
• Data analysts need to repeatedly perform a series of tedious and complex operations such as data collection, cleaning, modeling, and analysis.
• Business personnel are unable to work in a self – service manner, lacking operational tools. (Operational staff often do not possess professional data analysis skills. What they need are data usage scenarios closely related to the business.)
• Traditional data analysis methods usually lack an intuitive and visual interface, making it difficult to understand and apply the analysis results.
• The flexibility of the filter part in the visual interface is insufficient. It is difficult to search through hundreds of filtering conditions, and conditional combinations cannot be made as needed.

solution

• Establish a full – process service covering requirements, design, development, testing, and code. Through scientific project management methods, scientific agile development methods, and forms such as two periodic delivery regular meetings per week, help Xiaomi build a Web product that meets real – world requirements.
• Build a business intelligence visualization platform that meets the business requirements of customers by conducting Looker’s business model according to the Internet industry for customers, combined with Baidao’s industry experience, the business data structure of customers, and the needs of multiple business lines.
• The modeling and development team conducts in – depth analysis of Looker’s more than 30 categories and 429 API interfaces to provide support for a flexible semantic modeling layer for the front – end interface.
• The front – end development team develops customized Looker front – end interfaces with complex, rich, and dynamically interactive business content.
• The final product integrates the core of Looker’s powerful data analysis capabilities, as well as a flexible and feature – rich UI operation interface that is more in line with the usage habits of domestic users, making it a very powerful commercial cloud – native DA tool.

Use the product

• Firebase
• BigQuery
• Looker

Customer earnings

Customized Looker product is deployed to multiple product business lines such as Xiaomi Browser, Mi Gallery, and Mi Music. The comprehensive feedback from the business parties of each product line is as follows:

• Business personnel can directly use Looker for data exploration, which can replace 70 – 80% of the previous workload of running SQL to extract data.
• Data analysis conditions can be bookmarked, providing a historical query function. They have strong reusability, avoiding a large number of repetitive operations.
• Cloud – based web products are ready – to – use, enabling users to conveniently complete relevant work tasks either in the company or at home.
• It can be directly connected to the business data warehouse, providing high real – time performance for data analysis. At the same time, it can quickly query results for big data at the TB level.
• The modeling layer abstracts the data warehouse, which is very effective for business personnel to unify the indicator caliber. Moreover, it enables quick single – point changes to indicator fields and caliber terms, with the changes taking effect across the board.

Note: [All the following pictures are from the actual products of the customer]