How StarRocks Powers Data‑Driven Financial Marketing at Ping An Bank

Five‑in‑One Retail Banking Model

Ping An Bank’s retail‑finance strategy combines Open Banking, AI Banking, Remote Banking, Offline Banking and Comprehensive Banking into a data‑driven, AI‑centric operating model. Customer profiling, scenario analysis and internal resource orchestration are used to deliver precise services and generate a second‑growth curve.

From Product‑Centric to Data‑Driven Customer Service

The traditional product‑centric sales approach incurs high costs and low conversion. The bank adopts a "people‑goods‑place" framework based on three methods:

People (KYC) : Identify customers using identity verification and multi‑dimensional data analysis.

Goods (KYP) : Determine precise product and service needs through needs‑analysis.

Place (KYATO) : Select optimal delivery channels and scenarios for effective outreach.

Challenges in Financial Marketing Transformation

Financial institutions face complex user insights, strict compliance requirements, low‑frequency multi‑channel usage and high‑cost legacy processes, making data‑driven transformation difficult.

Smart Bank 3.0 Blueprint

Smart Bank 3.0 unifies the roles of operations managers, product managers and data analysts on a single portal. The architecture revolves around the three core elements (people, goods, place) and relies on StarRocks for enterprise‑wide analytical capabilities, including:

Real‑time behavior tracking and journey mapping.

Self‑service metric queries.

Customer‑data‑platform (CDP) tagging.

Strategy effectiveness analysis, A/B testing and API acceleration.

StarRocks in Financial Marketing Scenarios

StarRocks provides:

Channel & Strategy : Real‑time event back‑flow, multi‑table joins and multidimensional analysis for behavior data exploration.

Tag Management : Seamless integration of real‑time and offline data, eliminating the need for separate storage systems.

Strategy Platform : Fast evaluation of strategy brain, channel selection and customer reach.

A/B Experiments & Audience Analysis : Real‑time audience segmentation and metric generation, reducing latency from T+1 to seconds.

StarRocks‑Based CDP Platform

The CDP merges real‑time and offline data using StarRocks’ multi‑table JOIN, flexible tag filtering and bitmap operations. Key capabilities include:

Profile analysis and audience segmentation within seconds.

Materialized views simplify complex KYC grid calculations.

Exportable audience datasets for downstream analytics.

Metric Layer Architecture

The architecture is organized into three layers:

Physical Layer : Data engineers define an ER model with fact and dimension tables.

Logical Layer : A wide table aggregates all dimensions and metrics, enabling business users to drag‑and‑drop without understanding the underlying ER relationships.

Application Layer : Materialized DWS tables provide high‑performance queries for dashboards and self‑service analytics, achieving millisecond‑level response times.

This "define‑once, use‑multiple‑times" approach is supported by scheduled materialization to meet SLA requirements.

Indicator Scenario Design

StarRocks supports a three‑tier metric design:

Physical Layer : Build the ER model (facts, dimensions, objects) in the ODS.

Logical Layer : Transform the ER model into a super‑wide table that consolidates all customer dimensions and metrics.

Application Layer : Materialize frequently used metrics into DWS tables; DataAPI delivers millisecond queries, dashboards respond in seconds, and ad‑hoc analysis completes within tens of seconds.

Materialized views and bitmap operations enable rapid KYC calculations, tag saturation analysis and audience export.

Future Outlook

Ping An Bank plans to extend the StarRocks‑driven architecture enterprise‑wide, promoting transparent query rewrite, automatic materialization and cost‑effective big‑data solutions that make complex analytics as simple as using MySQL.