Creating A Custom ELT Process with Snowflake To Reduce Customer Churn
Creating A Custom ELT Process with Snowflake To Reduce Customer Churn
FirstLight designs, builds, and maintains fiber-based communication networks throughout the Northeast and has a network that spans over 25,000 miles of high-capacity fiber-optic cable throughout the Northeastern and mid-Atlantic United States and was looking to detect and prevent customer churn.
Approach
1
Gather
Archetype worked with FirstLight to examine processes and requirements, followed by creating and designing expectations for the development.
2
Review
Archetype reviewed FirstLight’s model design and created a proof of concept to get a faster review cycle in place. Additional details and efficiencies were examined that would have an effect on the models.
3
Test and Deploy
Before deploying a change or net new model, Archetype ensured all the requirements were first accounted for and that testing was sufficient. Models or reports were deployed at an agreed upon frequency (code always saved/updated).
Problem
With FirstLight’s business being dependent on monthly recurring customer revenue as a broadband and networking company, they were looking for more data and analytics around possible customer churn.
Solution
We implemented the following solutions for FirstLight:
Custom ELT (C# applications, API Code)
Snowflake
Reports created using Power BI
Snowflake data warehouse technology was used to organize and create an analytical reporting layer which allowed reporting out of Power BI. We also worked on a machine learning Dataiku workstream for some proof of concepts for propensity for customers to churn. Archetype created custom reporting asks from finance, marketing, and sales teams. Data sources have been added including custom API integrations for marketing sources.
Result
Through the integration with Snowflake, FirstLight was able to review data from their billing and customer service sources that provided insight into customer churn. The new process replaced a heavily manual process and allowed marketing analytics to provide insight around spend to create new networks.
