DataWind is an all-in-one data analysis and collaboration platform that supports self-service analysis of data at the hundred-billion scale. It enables seamless integration across the entire process, from data access, data integration, and query analysis to organization-wide collaboration and sharing. Through visualizations such as data portals, digital dashboards, and management cockpits, it empowers business users to achieve intelligent data insight and unlock the value of data.

The functional architecture of the DataWind product can be divided into the following six layers, viewed from the bottom up:

2.1 Data access #

Typically, the first step in a BI application is to complete data source integration. DataWind supports obtaining data from business databases, Excel/CSV files, business data entry on Lark, external platform data (such as advertising domains, content domains, and more), as well as real-time or offline data warehouses, currently supporting more than 40 types of data connections.

2.2 Query engine #

After data connections are established, analysis and computation on the application side require the support of an analysis engine. DataWind supports two modes: one is the built-in storage and computation engine ByteHouse, which enables self-service analysis of massive data at the hundred-billion scale. In most analytical scenarios (such as grouping, percentage, comparison, sorting, and so on), the proprietary ByteHouse query engine delivers at least a hundredfold increase in computation speed. The other is the direct connection engine, which interacts directly with the database and can be used when database performance is sufficient.

2.3 Data modeling #

After connecting to data sources, data can be loaded into the product by writing SQL or by directly loading raw tables into the product. However, these tables usually require some simple secondary processing to become datasets that can be analyzed.

• For data that has already undergone complex processing and only requires simple operations such as joining or unioning data, character type settings, or filtering, simple data processing and editing can be performed at this stage to create datasets for analysis.
• When data processing is highly complex, it is necessary to build an ETL pipeline. DataWind provides low-barrier AI+BI visual modeling capabilities for users with certain data processing skills, such as feature engineering, machine learning, and natural language processing.
  

2.4 Data analysis #

All operations in the data analysis process are performed through mouse clicks and visual drag-and-drop actions, and code or data scripts are automatically parsed into visual charts. Currently, there is a rich chart library, flexible query and analysis capabilities, as well as simple and user-friendly computational analysis features. At the same time, built-in attribution algorithms help users identify anomalies.

2.5 Data presentation #

For application scenarios, multi-platform applications can be implemented, supporting analytical dashboards, mobile devices, embedded in other applications, and large-screen displays. Meanwhile, it is deeply integrated with various common IM office products, such as Lark, DingTalk, and WeCom, enabling online data collaboration.

2.6 Data scenarios #

Data processed through the one-stop data analysis and collaboration platform can ultimately be used in multiple scenarios, including business decision analysis, market trend analysis, sales forecasting, growth analysis, inventory and sales analysis, anomaly diagnosis, channel analysis, and profile insights.

3.1 Product advantages #

(1) Easy to use: Zero threshold allows frontline staff to use it out of the box
DataWind provides a zero-threshold, visual operation interface, allowing each team in your company to easily complete data processing and build business data dashboards through drag-and-drop actions. According to different business scenarios, DataWind can also provide your company with DataWind solutions, enabling zero-threshold, fully self-service data analysis, and helping your company foster a data-driven culture for all employees. Currently, within ByteDance, DataWind supports self-service analytics for almost all business lines, covering the self-service analytics needs of the vast majority of employees. More than 200,000 active dashboards run daily, supporting over 5 million large-scale data queries, with daily active users exceeding 50,000. For some enterprise customers, daily active users have also reached several thousand or even over ten thousand.
(2) BI+AI: Large model capabilities and modeling attribution to uncover business truths
The analytics assistant provided by DataWind is built on large model capabilities, offering users an easy-to-use data visualization query and analysis solution. It enables core intelligent scenarios such as natural language-driven data insights and conversational automatic analysis guidance, reshaping data analysis workflows, further expanding enterprise data consumption, and promoting data democratization.
The visual modeling offered by DataWind allows business teams to construct, schedule, and execute complex ETL tasks through visual drag-and-drop operations. By integrating AI modules and data mining algorithms, complex data processing becomes simple.
The attribution analysis provided by DataWind uses built-in algorithms to achieve business data attribution with one click, automatically detects abnormal data, supports scenario-based analysis with automatic attribution and intelligent alert monitoring, and helps businesses uncover the truth behind data fluctuations.
(3) Fast collaboration: Office integration enables efficient data collaboration
DataWind is highly open, allowing close integration with enterprise applications to meet personalized enterprise needs. In particular, collaboration with online office tools enables you and your team to access data insights anytime and anywhere, assign responsibility to individuals for follow-up, and manage business risks. Specifically:

• Office integration for online collaborative work: DataWind can be integrated with various office tools used by your company, such as Lark, DingTalk, WeCom, and custom-built enterprise apps, to push data anomalies to relevant personnel or groups in real time or via monitoring alerts, ensuring anomalies are actively delivered to the right personnel or groups.
• Embedded analytics, seamless integration with enterprise platforms: DataWind can be seamlessly integrated into your company's proprietary platforms, providing embedded real-time data, reports, dashboards, and more, to enable interconnected data flows.
• Hierarchical data access control: DataWind can tailor an enterprise domain for each company, integrate with enterprise SSO, or provide a complete account system. Supports hierarchical access control for all administrative levels within the enterprise; supports hierarchical authorization management for different projects across various business lines and departments.

(4) Growth enablement: Synergistic product matrix to meet business growth needs
Based on the robust product matrix of BytePlus, DataWind can not only be combined with growth marketing products to implement marketing growth solutions such as private domain traffic operations, public domain traffic operations, and marketing growth analysis, but also form end-to-end data solutions with data middle office products. Specifically:

• We can build a digital operations platform for your enterprise through the three major products: DataWind, CDP, and GMP. By leveraging data analysis and insight, as well as refined audience segmentation, a more effective data-driven business operation system can be established, providing you with a digital platform that connects multiple data sources and builds an intelligent consumer growth system.
• We can gain in-depth data insights based on DataWind, identify potential issues, use product capabilities such as attribution analysis to find key groups, synchronize them to the CDP for targeted insights and refined operations, and finally track and analyze changes in key metrics based on CDP + DataWind to validate results, continuously iterate operational strategies, and truly achieve a closed loop from DataWind to decision-making and execution.
• We can closely integrate with data products such as DataLeap for data development and governance, and LAS for lakehouse analytics services, to build a one-stop data platform solution for you, covering data ingestion, data integration, data asset management, and data application.

3.2 Technical advantages #

(1) High performance: Second-level queries on hundreds of billions of data records
DataWind features a self-developed ultra-fast columnar OLAP engine, ByteHouse, created by an expert team, which solves the industry's performance challenges in big data multidimensional analysis, truly achieving second-level queries on hundreds of billions of data records and fully meeting the massive data application needs of large enterprises or rapidly growing companies. At the same time, DataWind adopts technologies such as multi-level data aggregation and hot-cold data tiering to further improve query performance and optimize storage costs.
(2) Big data: Highly scalable distributed cloud-native architecture
DataWind adopts a distributed cloud-native architecture with K8s containerized deployment and integrates big data components such as Spark, Flink, and HDFS, offering strong scalability to support large-scale data storage and computing. It is also highly open, enabling integration and interaction with existing enterprise data platforms in modules such as data permissions and task scheduling, achieving enterprise-wide data interconnectivity.
(3) Real-time: Enjoy real-time data import and updates
DataWind supports building real-time data pipelines based on Kafka, Flink, and ByteHouse. Based on ByteHouse's proprietary UniqMergeTree, it supports real-time data updates, improves data timeliness and visibility, helps enterprises perform real-time data analysis, and enables the creation of real-time data dashboards.
(4) Openness: Achieving data interconnectivity and meeting individual requirements
DataWind offers strong openness, providing comprehensive support for integration with and deployment on customers' existing systems, including embedded analytics, plugin market, data portal, office integration, and big data integration.

To quickly get started with DataWind, it is recommended to read User Learning Paths and begin using the product.