How Huolala Built Its Own Self-Service Data Analysis Platform from Scratch

Project Background

Business intelligence (BI) products are essential for most companies, enabling data‑driven decision making. Huolala, covering 352 Chinese cities with 660,000 active drivers and 9.5 million active users as of August 2022, faced increasing demand for rapid data flow and efficient analysis across its expanding business scenarios.

Project Goals

The self‑service data analysis platform, a core tool of the big‑data middle platform, must satisfy four key requirements:

Usability : Reduce the learning curve for employees with varying analytical skills.

Stability : Ensure sufficient compute capacity and smooth operation under high concurrency.

Scalability : Support a wide range of reporting scenarios and enable rapid reuse of report functions.

Security : Implement robust data‑permission mechanisms to protect core data assets.

Solution Selection

Is developing a BI product in‑house just reinventing the wheel?

Commercial BI solutions are abundant, ranging from traditional (SAP, IBM TM1) to agile (Tableau, PowerBI, FineBI) and cloud‑based tools (QuickBI, SugarBI). Gartner’s 2022 Magic Quadrant still places Microsoft, Salesforce (Tableau), and Qlik at the top, while domestic products are catching up.

A comparison of self‑development versus external procurement highlighted:

Requirement Support : Self‑development tailors features to specific business needs; external products offer generic functionality with limited customization.

Cost : Self‑development requires investment in product, R&D, testing, and servers (1‑2 years development); external solutions charge per system or per user.

Data Security : In‑house solutions keep data and personnel within the enterprise, reducing risk; external tools can be privatized but still pose leakage concerns.

Data Lineage : Self‑built platforms provide end‑to‑end traceability; many commercial tools lack complete lineage and report lifecycle features.

Requirement Response : Internal teams iterate quickly; external vendors involve longer support cycles.

Given the gaps in flexibility and security, Huolala chose to develop its own platform.

Practice and Exploration

How to plan a data analysis platform from 0 to 1?

The “Data Cloud Platform” aligns with commercial BI in offering offline/real‑time computation, an indicator library, and an OLAP engine for visual analysis. It reduces analysis barriers through drag‑and‑drop operations and supports five main modules:

Data Source Integration : Supports Excel, MySQL, Hive, Doris, and internal sources, with one‑click connections for internal systems.

Data Pre‑processing : Provides data modeling and custom SQL, data type settings, expressions, and field‑level permissions; multiple OLAP engines boost ad‑hoc query performance.

Data Visualization : Offers workbenches, chart components, dashboards; users can drag dimensions/metrics to create charts and switch chart types for diverse presentation needs.

Data Analysis : Enables fast metric calculation, multidimensional analysis, and statistical functions to extract key insights.

Application Scenarios : Includes metric exploration, reporting, large‑screen dashboards, and APIs, delivering end‑to‑end visual solutions for business data analysis.

These modules tightly interconnect, leveraging the big‑data middle platform’s capabilities to meet business data‑application requirements.

Issues and Challenges

How did we solve the risks and challenges?

Building a BI platform in‑house is inherently challenging due to extensive feature sets, high compute demands, and a broad user base. Post‑launch, functional bugs and feature gaps frequently arise, requiring a balance between needs and resources.

Comprehensive Planning : Conduct thorough requirement research, benchmark against leading BI products, and tailor designs to internal needs.

Flexible Solutions : Adopt a multi‑engine strategy to select the optimal compute engine for each data source, improving query efficiency and system stability.

Standard SOP Process : Classify users (developers, analysts, managers) and assign appropriate permissions; developers access the full stack, analysts use pre‑processed datasets, and managers view final applications.

Summary and Outlook

The article outlines the product design philosophy and technical choices for building a data analysis platform from scratch, offering a practical roadmap for similar initiatives. Although the Data Cloud Platform is still in internal incubation, it already serves multiple departments and is expected to cover about 80 % of Huolala’s data‑application scenarios, with future enhancements planned to narrow the gap with industry‑leading solutions.