Before the acquisition by Jabra, ListenLively operated as an audiology platform that offered a user-friendly online hearing test. Their goal was to empower individuals with accurate assessments of their hearing abilities and assist them in making informed decisions about hearing aid purchases. Users could take the test at home and consult with audiologists through video conferences. ListenLively facilitated the delivery of hearing aids directly to users, eliminating the need for in-person visits. Their platform aimed to provide comprehensive and convenient hearing care, improving the quality of life for individuals with hearing issues.
Problem
ListenLively ingested data from two sources: Amplitude and Salesforce. The data in Salesforce was disorganized, not documented, difficult to access, and use. ListenLively needed help in integrating the two data sources together for better insights. More importantly, they wanted help in cleaning the data and building dimensional and fact tables for business intelligence.
Our goal was to help ListenLively build a data model within the existing data warehouse and develop pipelines to transform existing data into clean datasets using dbt, onto which a BI tool can easily be mounted.
Approach
Data Sleek broke down the project into three phases: DBT Cloud Setup, Dimensional Modeling, and Data Quality Control.
DBT Cloud Setup
The objective of this phase was to setup DBT Cloud to manage the DBT CI/CD pipeline. DBT Cloud is the hosted option of DBT. It handles all of the scheduling, dbt code source control, and has a UI to interact with.
Dimensional tables provide descriptive attributes and context to the data stored in the fact tables. They capture various dimensions by which the data can be analyzed. Each dimension table represents a specific aspect of the data, such as time, product, location, or customer. Dimensional tables typically have a primary key column that uniquely identifies each dimension record, and additional columns that contain descriptive attributes or hierarchies.
Fact tables, on the other hand, store quantitative and numerical data representing business events or transactions. Each row in the fact table corresponds to a specific instance of the event or transaction, and the columns contain the measurements associated with the events. Fact tables have foreign keys that link to the corresponding dimension tables, establishing relationships and providing context to the data.
For this project, Data-Sleek took three steps to build each fact and dimension table:
Explore the existing tables for table attribute discovery. This includes updating the data dictionary.
Build the Dim/Fact tables ERD and consult with ListenLively for approval.
Implement the Dim/Fact tables in DBT, and test them using DBT runs to make sure that the tables are properly populated and do not contain anomalies (like missing or truncated data).
Data Quality Control
During the Data Quality Control phase, the objective was to ensure that the output from the fact tables aligns with the expected results. ListenLively was involved in this phase to validate that the results fall within acceptable ranges. Once the results were deemed satisfactory, the DBT code was committed to the Master Repository, allowing for potential updates if errors were identified at a later stage.
Results
Data-Sleek created fact tables which the business intelligence team at ListenLively was able to use to track target KPIs. Non only fact and dimension tables significantly improved ListenLively’s query performance it allowed ListenLively to slice, pivot data in many ways by joining few tables. Building fact and dimension tables.
By combining the data from fact and dimension tables, analysts can track specific KPIs by querying and analyzing the data based on the desired dimensions, metrics, and levels of granularity. The fact tables provide the quantitative measurements, while the dimension tables provide the context and segmentation options necessary for effective KPI tracking and analysis.
Results
Health Karma’s partnership with Data Sleek resulted in a finely tuned system for tracking the performance of enrollment procedures. Now there are efficient data flows in place that enable operational data, such as sending registered user’s data to Mailchimp, for continued customer engagement. The dashboarding system in place allows business stakeholders to gain insights independently, using simple queries, yielding elegant summary data and visualizations.
