Enabled a pharmaceutical company to increase their Conversion Rates, decrease Turn Around Time (TAT) and helped them realize the Return on Investment (ROI) on marketing cost.
Revolutionizing Hispanic Health with Data Analytics
About the Client
Our client is backed by leading retail partners, and leading medical professionals, and is one of the leading pharmacy and healthcare companies headquartered in the United States.
The Business Challenge
The client had troubles in identifying the resources needed for the call center team based on the campaigns, seasons and time-period of the day.
The Business Approach
Forecast calls at a 15 minutes interval and resources required for those calls to improve the calls to answered ratio
Data Preparation
Data from various sources were fetched such as
- Call Data
- Yearly Calendar
- Weather
- Staff Data
- Created structured tables such as having the time stamp of the call date, call ID’s if it’s an inbound or outbound call and the call duration and the ringing duration of the call to separate the actual call time.
- Gave a flag if that particular day was a holiday
- Gave another flag if it was a sunny, rainy or a snowy day
- Aggregated the data for every 15 minutes for that day
Tech Stack
The tech stack used in this project are
- AWS Databricks python (create the model and dataset)
- Snowflake (store the dataset)
- AWS S3 (store static master files)
- Google calendar (calendar and holiday data)
- Python-weather (take the weather details)
Modeling Approach
- Compiled all the data and then gave that as an input to the model. Since the data we had was non-linear, w negated the linear models
- Three models were used LSTM, XGBoost and Random Forest
- While trying with the above three models, we observed XGBoost gave accurate results with the data with testing data to be as accurate as 86% and with 20-80 train test principle
- Following calls prediction, Erlang method was used to forecast the resource required to answer the calls
Model Results
- With the initial training, the model gave 86% accuracy.
- When tested in real time, the model gave 82-85% accuracy in the initial two weeks and after 2 retraining in a month, we extended the model accuracy till 90%
Impact Created In Business
- This helped Call Center heads to pre-plan their resources and ensured they are not understaffed and at the same time not overstaffed as well
- Improved the productivity as answered call went from 72% to 85%
- Conversions improved from 45% to 68%
Reference Architecture
Related Case Studies
Achieving low-latency API-based queries with Mongo DB
Revolutionizing Pharma Analytics with AWS Data Lake
Boosting Performance with Apache Spark Migration
Unlock the Potential of Data Science with Aptus Data Labs
Don't wait to harness the power of data science - contact Aptus Data Labs today and start seeing results.
If you’re looking to take your business to the next level with data science, we invite you to contact us today to schedule a consultation. Our team will work with you to assess your current data landscape and develop a customized solution that will help you gain valuable insights and drive growth.
