Product Sales Remaining Opportunity
BI Solutions
November 2015
Objectives
- Form a ranked list of ZIP codes from states with reasonable remaining product sales opportunity and determine ZIPs that have the highest expectation of additional generation of revenue
- Estimate cost-effectiveness of product in-person promotion to generate additional sales in selected ZIPs
Findings / Conclusion
- Selected 15 states with reasonable remaining opportunity: AL, CA, FL, IL, KY, MO, MS, NC, NY, OH, OK, SC, TN, UT, VA can produce additional 4 MM units (57%) sales
- Selected fifteen hundred ZIP’s (about 20%) can generate additional 2 MM units sales (50%) of remaining opportunity from selected states
- Available data does not support in-person promotion as cost-efficient/effective promotion channel
- Non-in-person promotion is recommended
- Controlling for all predictors (ZIP Level Projected number of customers, ZIP Level Number of Units Sold in Pre Period, ZIP Level Number of in-person Promotions, State, and ZIP Level Remaining Opportunity), the following states:
- Alabama, Kentucky, Mississippi, Oklahoma, and Tennessee have the highest propensity of increasing sales revenue
- Florida, Missouri, New York, Ohio, and South Carolina are the next group of states
- California, Illinois, North Carolina, Utah, and Virginia have the lowest propensity of increasing sales revenue
Number of ZIPs by State
(Data generated graph)


ZIP Level Response Curve
Method: Stochastic Gradient Boosting
Model: Non-Parametric Robust Data Mining Regression
- Dependent Variable: ZIP Level Unit Change
ZIP Level Unit Change = ZIP Level Unit Count in Current Period
- ZIP Level Unit Count in Previous Period
- Current period: recent 12 month
- Previous period: previous 12 month
- Predictors (in order of importance on 100-point scale):
- ZIP Level Projected Number of Customers (Score = 100)
- ZIP Level Unit Count in Previous Period (Score = 69)
- Number of ZIP Level In-person Promotions (Score = 41)
- State (Score = 22)
- ZIP Level Remaining Opportunity (in units) (Score = 14)
ZIP Level Unit Change Response Curve, controlling for all other predictors in the model
(Model generated graph)
Impact of State on ZIP Level Unit Change, controlling for all other predictors in the model
(Model generated graph)