Case Study 14: Developed an NBA model that recommends the next best actions based on HCP behavior

Business Problem

  • Reps struggled to effectively correlate information from various reports and systems. 

  • Unable to adjust messaging as the patients were switching dynamically.

ProcDNA Approach

Raw Data From Source

Identified the right data sources (example – Weekly/Monthly Sales, Call activity, Email engagement, Website, Speaker Bureau etc.) for each suggestion and model training.

Constraint File

Incorporated constraints like rep time-off, call/email capping, HCP consent, etc.

Model Training

Trained the deep learning model (TCN & FCN) using the past 6 months data like Sales, Call, Email & HCP affinity.


An intelligent optimization algorithm – Monte Carlo Tree Search was used to minimize the number of iterations needed to identify the promotion sequence that returns the highest value for an HCP. 

Veeva Suggestions

Predicted alerts were sent as suggestions on Veeva app using Veeva API integration. 

Client Impact

IVA Call Utilization

IVA call utilization increased from 75% to 90% post suggestions launch

Increased Product Sales

Based on control test analysis, product sales increased by ~4% within the first year of suggestions launch

Meaningful Conversation

Sales reps were able to have meaningful conversations with HCPs


Let’s Get To Work