THE PROBLEM
Client, a Fortune-500 FMCG manufacturer, provides discounts to the retailers as part of its B2B email campaigns. However, the client wanted to improve the ROI on the discounts offered. Hence, they partnered with Inxite Out with an objective to decrease the discount expenditure while maintaining or improving the sales figures.
INXITE OUT APPROACH
Data Normalization
The client sells myriad of brands and SKUs through its vast retailer network. There are several different modes and methods of providing discounts too. The very first step of our solution involved understanding of this business context, followed by cleaning and normalization of the discounts data in a centralized data lake to make them amenable for further processing downstream.
Broad-based Discount Sensitivity Profiling
Implemented custom algorithm to classify the retailers broadly into Discount Takers and Discount Non-takers. The Discount Takers were further categorized into three different classes (Highly Sensitive, Moderately Sensitive, Not Sensitive) according to their sales uplifts or responsiveness to changes in discounts. An approximate local time series similarity analysis on past purchase behavior along with tunable custom thresholds were used for this purpose.
Feature Enrichment and Model Development
Curated and enriched the data further with factors such as campaign information, discounts offered during the current / last / next campaign, brand involved, past purchase behavior of the retailer etc. Leveraged AI to fine-tune the predicted sales uplift of any given retailer by using a non-linear model that is aware of retailers’ present needs and business context.
Optimized Recommendations
Finally, the above predictions were converted into recommendations for discounts to be offered for optimal ROI, while allowing the client to incorporate specific business constraints where necessary. Model recommendations were validated by running on-field campaigns on pilot market.
RESULT
A/B testing on 500+ accounts for target brands in the pilot market during the test period yielded ~10% cost savings across different retailer clusters, while also maintaining >1% sales uplift.