Implementing Full Stack Data Solutions for Soundstripe

Opportunity

Soundstripe helps creators and businesses find the perfect music and sound effects for their videos with millions of songs and SFX to choose from. As their business grew so did the amount of data that they were collecting on core product interactions like downloading and favoriting, browsing and traffic behavior, as well as marketing spend and activity on third-party platforms.

While they had portions of data being collected and reported on they had some key challenges they wanted addressed:

• ‍High Data Dispersion - data was siloed and separated making integrated insights a challenge.
• ‍No Data Dictionary - data that was reported on might have contradictions depending on what team was compiling the report.
• ‍Slow Time-to-Insight - some key reports and KPIs used to make business decisions took weeks or a month to compile.
• ‍Low Adoption of BI Tools - existing business intelligence tools had low adoption rates among employees.

Soundstripe partnered with Alchemy Collective to develop and implement a data strategy and data stack that would address these challenges and unlock future value.

‍

Our Approach

We rebuilt Soundstripe’s data stack from the ground up beginning with an efficient extract, load, and transform (ELT) schedule, establishing a verified table schema for their data warehouse, implementing a new BI tool for dashboards and reports, and developing predictive models like time-series demand forecasting and customer lifetime value leading to significant reductions in time to insight and enabling teams to automate or make data-driven decisions faster.

Extract and Load the Raw Data to Snowflake

• Utilize data integration tools and custom scripts to extract raw data from various sources including first-party databases, product analytics platforms, advertising platforms, and APIs.
• Load the extracted data with batch ID markers and load date timestamps into designated raw input tables in Snowflake staged for downstream processing and transformations.
• Optimize data replication schedules and procedures for faster overall completion of extract and load process (i.e. hourly or daily, drop/replace or incremental).
• Setup automated Slack and email alerts in order to monitor performance and error handling.

‍

Define the Desired Database Schema

• Design a well-structured database schema that aligns with the business requirements and analytics needs.
• Determine the appropriate data types, constraints, and relationships among tables.
• Create Snowflake analytics tables based on the defined schema.
• Implement data governance policies to enforce data standards and ensure compliance.

‍

Transform the Data Using dbt

• Leverage dbt (data build tool) to automate data transformation processes.
• Write modular SQL scripts (models) to perform data cleaning, transformation, and aggregation.
• Leverage dbt's testing framework to validate the accuracy and consistency of data transformations.
• Orchestrate data transformations using dbt's dependency management and build pipelines.
• Integrate dbt with Snowflake to seamlessly execute transformations within the Snowflake environment.

‍

Develop Prioritized Dashboards in Sigma

• Identify key business metrics and develop interactive dashboards in Sigma, a business intelligence tool, for downstream consumption.
• Design dashboards that provide insights into various aspects of the business, such as revenue performance, user behavior, and content engagement.
• Implement drill-down capabilities to allow users to explore data at different levels of granularity and abstraction.
• Host office hours and webinars to increase adoption within the company.

‍

Train and Deploy Demand Forecasting Model

• Collect historical traffic and purchase data, product information, and relevant market factors.
• Choose an appropriate model that will balance speed, accuracy and robustness.
• Train the ML model using the selected algorithm and historical data.
• Evaluate the model's performance using metrics like mean absolute error (MAE) and root mean squared error (RMSE).
• Deploy the trained model into production using Snowflake Machine Learning.

‍

The Results

Over the course of our engagement we helped Soundstripe:

• Centralize priority data in a single data warehouse environment.
• Standardize KPIs and data transformations.
• Reduce time-to-insight from weeks to hours on key metrics.
• Increase adoption of business intelligence tools by 5X.
• Automate CLTV predictions and setup real-time feedback to third-party platforms for optimization purposes.