Design and Implementation of the JiShi BI Data Visualization Platform

In the context of digital transformation, data has become a critical foundation for enterprise development, and effective data analysis is a primary challenge for many companies.

The internally developed JiShi BI platform builds on traditional BI multi‑source data management, offering self‑service data display page generation and enhanced analytical capabilities such as anomaly, conversion, and funnel analysis, which improve analyst efficiency and enable deeper business insights.

JiShi BI differs from conventional BI tools by providing a complete data closed‑loop: data extraction, cleaning, analysis, and report output, thereby supporting enterprise decision‑making.

Core Process (Section 3.1)

Data Connection: Users connect the tables they need to query into the BI system.

Data Preparation: Query and organize tables to form datasets with annotations.

Data Dashboard: Configure dashboards with the corresponding datasets, drag‑and‑drop components, and solidify them as visual boards.

Data Application: Configure application information to expose dashboards externally.

Enhanced Analysis: Apply algorithmic models for advanced analysis such as anomaly detection and conversion analysis.

Architecture (Section 4)

The platform follows a four‑layer modular design:

Data Access Layer: Unified data source management with permission control, supporting real‑time validation and synchronization.

Data Service Layer: Custom query engine handling diverse data sources (MySQL, ClickHouse, Doris) to reduce development cost and provide parsing and aggregation services.

Data Presentation Layer: Builds user‑defined analytical dashboards, manages permissions, and supports multi‑device rendering.

Enhanced Analysis Layer: Integrates algorithmic models for intelligent diagnostics, dimension drill‑down, and metric exploration.

Detailed Designs

Data Access Layer emphasizes unified data sources, permission‑based management, and real‑time validation.

Data Service Layer addresses challenges of heterogeneous query engines by developing a unified query engine and aggregation service.

Data Presentation Layer includes dashboard construction with component optimization, interaction improvements, and flexible field filtering.

Application Management allows configuration of unique authorization keys, domain settings, and token lifetimes for secure third‑party integration.

Enhanced Analysis

1) Anomaly Intelligent Analysis uses the Adtributor algorithm, calculating surprise (S) and explanatory power (EP) to identify abnormal dimensions and provide attribution.

2) Funnel Strengthened Analysis supports ordered and unordered funnels, offering conversion metrics, loss diagnosis, and visualizations via ClickHouse’s windowFunnel function.

Conclusion (Section 5)

JiShi BI extends traditional BI by adding dashboard customization and AI‑driven analytics, already applied to monthly advertising reports and operational channel performance, improving efficiency, deepening data insight, and providing timely alerts for decision‑making. Future enhancements will focus on real‑time analytics, intelligent reporting, and proactive monitoring.