How BMW Turned Data Into Growth: A Sensors Data Case Study

Background: Digital Transformation Imperative

BMW's digital story began with the MyBMW and MyMINI apps launched in China in 2020, which added community, membership, and e‑commerce modules to create a social brand experience and boost user engagement.

By the end of 2021, the two apps reached over 3 million active users and 1 million monthly active users (MAU); by March 2023, active users surpassed 10 million with MAU over 2 million. This rapid growth exposed the limits of the existing experience‑driven operation model.

Core Challenges: Data Collection for a Global Automaker

BMW faced several obstacles:

Global standards vs. local needs : The headquarters mandated a strict event‑tracking specification that conflicted with Sensors Data’s schema, requiring extensive mapping work.

Multiple tracking solutions : Simultaneous use of Countly, MZ, and Sensors Data created data silos and format incompatibilities.

Complex engineering effort : Integrating Sensors Data SDK with the Flutter‑based app demanded a custom compatibility layer.

Documentation overload : Maintaining four separate tracking documents (design, mapping, map, version) became error‑prone.

Massive historical data migration : Over 100 TB of legacy event data needed to be imported without breaking analytics continuity.

Solution: Systematic Engineering from Specification to Platform

The team built a three‑stage solution:

Standardization : Unified tracking design for more than 1 000 events, created a mapping layer to translate Countly/MZ events to Sensors Data, and introduced a “tracking map” documenting over 100 core pages.

Platformization : Developed a centralized Tracking Management Platform to store design, mapping, version, and map artifacts, eliminating manual document maintenance.

Engineering implementation : Implemented a preprocessing plugin in Sensors Data’s SDF module to convert incoming raw events to the Sensors Data EUI format using mapping metadata stored in MySQL. Failed mappings generate alerts and dirty‑data files for later correction.

Historical data migration used a multi‑phase strategy: collect legacy data, run parallel pipelines with Sensors Data, then decommission the old system. The pipeline pulled Parquet files from OSS, flattened rows to JSON, enriched records with dimension data from the data‑warehouse (SKV), performed attribute validation, and finally batch‑imported into Sensors Data.

Project Outcomes: Data‑Driven Business Growth

Key results include:

Over 300 operational dashboards built on Sensors Data, enabling fine‑grained decision‑making.

Real‑time recommendation engine powered by behavior data increased community content click‑through rate by 301%.

The Tracking Management Platform provided an end‑to‑end workflow (requirement → design → review → testing → release) for all tracking needs.

Future Outlook: Towards an AI‑Powered “Smart” Era

BMW plans to upgrade Sensors Data from version 2.x to 3.0.2 and explore AI integration. An envisioned “Agent” will allow business users to ask natural‑language questions about tracking definitions, data differences, or similarity between features, leveraging large language models to deliver instant, data‑backed answers.

By constructing a closed loop—from tracking documentation and map, through the management platform, to Sensors Data and the CDP—BMW aims to make data more usable and less burdensome for future product teams.