AI Data Analytics Platform for the Music Industry
A cloud analytics platform that brings together data from about 40 distribution platforms, about 100,000 records a day. Executives and marketers ask questions in plain language and get the analysis back.
Challenge
A Japanese music industry company reaches listeners through about 40 distribution platforms (aggregators). Each platform sends back its own distribution data in its own format, adding about 100,000 new records every day. Analyzing sales performance or understanding the cost structure used to require a dedicated engineer to write SQL queries every time. The data was scattered, and management had never truly had a "bird's-eye view" of the overall trend across all platforms.
Solution
Scope of work: Architecture design and rebuild of the data analytics platform, plus development of a natural language query system.
- Data pipeline: Raw distribution data from every platform lands in S3. Lambda cleans it and normalizes the formats, and SQS and Snowpipe write it continuously into Snowflake, where unified master tables are built in SQL inside the data warehouse.
- Visualization layer: Analytics dashboards for executives and marketers built with Streamlit, with the most-used dimensions ready out of the box.
- Natural language queries: A query system on top of the data warehouse that takes questions in plain language and returns the analysis directly, so business users who do not write SQL can pull data on their own.
- Delivery timeline: About one year, covering the full chain from data ingestion to dashboard launch.
Results
- Cross-platform trends visible for the first time: Data from about 40 distribution platforms is aligned in a single table, and management can now see the overall trend across platforms for the first time.
- No more waiting in line for engineers: Executives and marketers pull data directly from the dashboards or in plain language, freeing technical staff from repetitive data requests.
- Reusable for multi-source data: The "heterogeneous multi-source data → cloud data warehouse → dashboards and natural language queries" architecture applies equally to unified analysis of multi-channel sales data or production line data across multiple factories.
What the client said
"Just asking a question in plain language is enough to run an analysis. Insights are far more within reach, and we depend much less on specialist engineers."
See other case studies
Eight platform-wide sales events in over two years with zero major incidents, load tested at up to 8 times normal traffic.
Video data and transaction data connected across about 500 stores, cutting the investigation of a suspicious transaction from a week to a day.
