How a CPG Leader Reduced Discount Costs While Simultaneously Increasing Sales 

Client Context  

Our client, a Fortune 500 CPG manufacturer, used B2B email campaigns to send targeted discount offers to their network of retailers. 

This process created significant cost leakage, as the same offers were rolled out to all retailers without a data-driven method to determine which ones actually needed a discount to purchase versus those who would have bought anyway. 

The client partnered with InXiteOut to develop a targeted approach to optimize discount expenditure while generating maximum sales uplift. 

The InXiteOut Approach 

To meet the client's objective, InXiteOut implemented a four-stage, AI-driven solution. 

Data harmonization and 360-degree view 

The client's data, spanning multiple brands, SKUs, and discount methods, was first centralized and harmonized in a data lake. This created a 360-degree view of each retailer by combining data from: 

• Sales and Purchase History: Past transactions, order frequency, and volume. 
• Discount Data: Historical offers and redemption patterns. 
• Campaign Data: B2B email campaign engagement and offer details. 

Retailer profiling   

Using the unified dataset, we built a two-part retailer profiling module: 

• Discount Sensitivity Profiling: A custom algorithm analysed each retailer's past purchase behaviour to classify them into segments, such as: 
• Discount Non-Takers: Retailers who purchase consistently, with or without a discount. 
• Discount-Sensitive: Retailers whose purchasing volume directly increases with discounts. 
• Discount Indifferent: Retailers who use discounts but show no significant sales uplift. 
• Sales Uplift Prediction: For the key "Discount-Sensitive" segment, a predictive AI model was built to forecast the specific sales uplift a given retailer would generate at different discount levels. 

Developing the recommendation engine 

Leveraging the Profiling and Uplift Prediction model, we developed a discount recommendation engine. The engine was designed to determine the optimal discount to offer each retailer to maximize ROI. 

Pilot, validation, and productionization 

The engine's recommendations were first validated in a pilot market through a live A/B test. This test compared the new AI-driven recommendations against the client's previous discount strategy to measure the impact. 

Following the successful pilot, the solution was productionized, creating an automated pipeline to continuously update the models and recommendations and integrate with the campaign engine for automated distribution of offers to retailers.  

Technology stack used 

• Azure ETL and ML Ecosystem (Data Lake, Data Factory, Databricks)     
• Power BI  

Benefits Delivered  

The solution was validated over a 6-month period, delivering consistent results that achieved both of the client's primary objectives: 

• 12% Reduction in Discount Costs: The model delivered a 12% cost saving by precisely targeting discounts, reducing wasted spend on "Non-Taker" and "Indifferent" retailer segments. 
• ~3% Increase in Overall Sales Uplift: The highly targeted offers to "Discount-Sensitive" retailers led to a measurable increase in overall sales.