Building a Multi-Dimensional Analysis System: Practice at Baixin Bank

The speaker, the BI head of Baixin Bank, shares the bank's business background and the motivations for establishing a multi-dimensional analysis framework to support its fully online, internet‑only operations.

Baixin Bank, a state‑backed direct‑sales bank launched in 2017, serves a 2C+2B customer base with consumer and industrial finance products, requiring comprehensive data to drive efficient online decision‑making.

The data analysis demands are distilled into three principles: accuracy (统一业财口径), completeness (covering core scenarios across the full data chain), and ease (providing differentiated views for strategic, tactical, and operational users).

To meet these, a hierarchical indicator system is built, exemplified by a credit‑business scenario: high‑level business goals are broken down into KPI groups, then into primary, secondary, and tertiary metrics, with both generic and analytical dimensions defined for each level.

The analysis system follows a staged approach: descriptive analysis to answer “what happened?”, diagnostic analysis for “why it happened?”, predictive analysis combining expert insight and algorithms for “what will happen?”, segmentation‑driven iteration for “what is happening?”, and finally decision‑support analytics to empower business actions.

Data products are constructed using a three‑layer, four‑block cube model (enterprise‑product‑user) that enables drill‑down from high‑level dashboards to product‑level diagnostics and user‑level behavior chains, supported by self‑service drag‑and‑drop visualizations, automated alerts, and attribution analysis.

The concluding reflection presents a pyramid‑plus‑pentagon framework where data breadth (coverage) and depth (business‑technology integration) guide the evolution from data modeling and assetization, through visualization and operational analysis, to decision‑support layers, emphasizing practical, high‑impact data solutions.