MARKETING ANALYTICS
GoodBelly Market Response Modeling
This was an individual assignment completed for Marketing Analytics I in order to understand the effect of promotional advertising on demand/sales.
Assignment
Next Foods, Inc. based in Colorado, produces a range of probiotic juice products under the name "GoodBelly". These products were introduced to the market in January 2008 and are available nationwide at various retailers including Whole Foods Market and Safeway, Inc. The company invested in in-store product demonstrations between May and July 2010 in certain regions of Whole Foods. GoodBelly is interested in analyzing the effect of these demonstrations on sales and profitability.
To investigate this issue, GoodBelly has collected data from two regions: Rocky Mountain (RM) and Northeast (NE). This data includes information such as units sold, retail price, a binary code for demonstration (with a value of 1 if a demo was held at a specific store in each region during a particular week), and a dummy code for demo1-3 (with a value of 1 if a store had a demonstration held 1, 2, or 3 weeks ago).
Analysis
For each region, RM and NE, I first estimated the sales response models separately by specifying sales as a linear function of price, demo and demo1-3.
I found that as price increases, sales decreases. Consumers are more price sensitive in the RM region than the NE region. More specifically, in the RM region, when price increases by 1%, sales decreases by 76.99%. In the NE region, when price increases by 1%, sales decreases by 36.20%.
When a store has a demo or had a demo in the last three weeks, sales increase. Demos have a larger positive effect on sales in the RM region. More specifically, in the RM region, if there was a demo in store that week, sales increased by 130.66%. Additionally in RM region, if there was a demo in store 1-3 weeks ago, sales increased by 89.45%. In the NE region, if there was a demo in store that week, sales increased by 107.78%. Additionally in NE region, if there was a demo in store 1-3 weeks ago, sales increased by 63.79%.
Next, I estimated a model that pools the RM and NE regions into a single model with a common intercept, price coefficient, demo coefficient and demo1-3 coefficient.
The pooled model demonstrated that when a store has a demo or had a demo in the last three weeks, sales increase. More specifically, if there was a demo in store that week, sales increased by 136.27%. If there was a demo in store 1-3 weeks ago, sales increased by 87.93%.
Now the question is, are these segments (Region RM & Region NE) statistically different? In order to answer that question, I created interaction variables to determine whether or not the price coefficient and demo/demo1-3 coefficients are statistically different across regions. I looked at four models:
Different intercepts only
Different intercepts and price coefficients
Different intercepts and demo/demo1-3 coefficients
Different intercept, price and demo/demo1-3 coefficients.
Conclusion
Based on my findings, I choose Model 2 and used the parameters to test for an effect of running demos in all stores in the RM and NE markets for the week of July 20,2010. This model includes any dynamic effects of the demonstrations estimated to be present in the data. I assumed that the retail price in each store will remain at July 13, 2010 weekly prices for the week of July 20, 2010 and any subsequent weeks relevant to your analysis.
Assuming retail % margins of 30% and manufacturer % margins of 50% (constant across regions, stores and weeks), I found expected profits for GoodBelly. The expected profits were $39,175.77.
In order to evaluate the importance of the dynamic demo effect in the weeks subsequent to a promotion, I re-estimated the model without the lagged variable. The dynamic demo effects are relatively important in the weeks subsequent to a promotion. The aggregate profit estimate (total for all four weeks) would be $4,551.04 lower WITHOUT the dynamic/lagged effects.