How Big Data Powered a Dance App’s Explosive Growth: A Real‑World Case Study

Challenge

When the Tangdou app for square dancing launched, the team faced unknown user profiles, frequent crashes, unclear growth direction and a lack of data‑driven decision making. They needed to know how to acquire users, which songs were popular, usage paths, crash frequency, and more.

Opportunity

The team decided to use big‑data technology to empower the business, evaluating two options: building an in‑house BI platform or purchasing a data‑analysis service.

Build BI

Advantages: Fast response and the possibility of creating a unique analysis system.

Disadvantages: Requires hiring data analysts and big‑data ops, low efficiency in data request cycles, complex ETL work, and repeated development for common metrics.

Buy Data Service

Advantages: Quick impact and cost savings on BI staff.

Disadvantages: Ongoing service fees and integration effort.

The company chose the professional analytics provider Sensors Data (神策数据).

Core Demands

Operational decisions – daily active users, retention, geographic and device distribution, search clicks, push reach and open rates.

Quality optimization – page view counts, tab exposure, crash root‑cause analysis.

Private‑cloud deployment – data must stay inside the IDC and support 2‑3 years of capacity without bottlenecks.

Solution

Sensors Data supplied a digital‑operation closed‑loop solution, embedded an SDK for event collection, and provided multi‑dimensional analysis without predefined metrics.

Visualization of basic retention metrics:

Another metric overview:

Deployment & Performance Testing

Two deployment options were evaluated on UCloud hosts: SSD cloud‑disk VMs and Kuaijie AMD VMs, both with 32‑core CPUs, 64 GB RAM and 2 TB storage.

The Impala‑TPC‑DS benchmark suite was run on 500 GB of data to compare query latency.

Architecture diagram:

Results showed the Kuaijie AMD host improved query performance by 44.98 % while keeping storage cost unchanged.

Performance gains were attributed to NVMe SSDs with RDMA, 25 Gbps networking, and second‑generation EPYC processors.

Outcome

The Tangdou app now handles massive daily data reporting, analysis and statistics, meeting both current and future performance and capacity requirements.