SEI Discount Program
Profitability Analysis
Kan Ito
July 24, 2017
Overview
Analysis based on transactional data from 2015-2016 on the following:
Bogota($75), South Face ($85), Pangea ($105) brands
Sold in, Jacksonville, Albany, Springfield, Eugene, Bend, and Tacoma
Data includes
All transactions with quantities of jackets sold
Price of all jackets per day
National Discount Program schedule (seasonal & holidays)
Objective: measure effectivity of discount program and recommend future
discount programs for short and long term.
Initial Assumptions
Define profitability as the sales of jackets
Each transaction is limited to only one type of jacket, implied by the data
Sales location and brands not correlated with each other.
Location and brand variables are independent of each other.
Empty data cells in daily prices were assumed to be at retail price with the
respective discount.
I. The Data
Seasonal variation
Other cities display the same sinusoidal pattern where sales are
unsurprisingly correlated with time of year.
Discount Effect
Similarly, data extracted to find correlation for sales and
discounts. Initially, this data shows weak signs of the correlation
between the two.
Brand Effects
The brand categories have also proved significant signs of
influence in the sales of jackets. Below by brand,
by location…
II. Fitting a Model
Multiple Regression
Data suggests that seasons, cities, and brands are all significant variables
that affects the sales. With several known variables, multiple regression
is a suitable choice to model the sales. Furthermore, discount price will
be added to the list of variables since this is our primary interest.
Initial Two Variable Regression
Sales = 27.1 + 0.91 * Discount + -14.5 * season
Coefficient of Determination: R-squared = 0.79
As expected, the coefficient for the discount is much smaller than that of
seasons. This implies a much stronger correlation for seasonal changes in sales.
*Note that Discount is scaled by a factor of 10 and season is discretized into 365/2~183 slots for effective optimization
Complete Multiple Regression
Independent Variables: (15 total)
• constant • discount^2
• discount • season^2
• season • bool constants for 6 cities
• discount * season • bool constants for 3 brands
Complete Multiple Regression
Coefficients
Constant 32.42 Springfield 12.40
Discount 4.24 Eugene 4.92
Season -30.50 Bend 5.90
Discount*Season -1.15 Tacoma 5.67
Discount^2 -0.15 Bogota 0.92
Season^2 -1.87 South Face 26.2
Jacksonville -5.63 Pangea 5.19
Albany 12.90
Coefficient of Determination: R-squared = 0.81
III. Recommendation
Discounts incur larger sales variation in colder seasons. Therefore, the discount program
should be utilized more in colder seasons, and less in warmer seasons. Separate simple
regressions were conducted to verify such phenomena.
Discount Coef: 0.71
Discount Coef: 0.33
Colder season
Warmer season
Similarly, the brands and locations can be utilized more from the discounts. Pangea
displays the highest sensitivity to discounts. South Face is second, and Bogota shows very
little sensitivity. Therefore, the discount program should be utilized more for Pangea,
then South Face in this order of preference.
For location, the sensitivity to discounts was much less varied. The locations of
preference for discount programs are: Albany, Tacoma, Springfield, Eugene, Bend,
Jacksonville.
The coefficients for the separate regression results are below:
by brand: Brand Discount Coeff
Bogota 0.71
South Face 5.22
Pangea 8.99
by location:
Location Discount Coeff
Jacksonville 4.38
Albany 8.44
Springfield 6.18
Eugene 5.90
Bend 4.41
Tacoma 6.27
IV. Conclusion
Objective: provide recommendations for discount program to increase profit
Data suggests the strongest correlation with seasons. Discounts have a
relatively weak correlation.
Model of choice: Multiple Regression with essential independent parameters
discount, season, cities? Brands?
The multiple regression model revealed variation in product of variables with
discounts and several brands and cities in correlation with discounts. The
variation in this correlation revealed that discounts should rather be
implemented for winter seasons as opposed to summer.
Similarly, the discounts should be emphasized for sales for the Pangea brand
and locations in Tacoma and Albany.
