Effect of promotions on sales and classifying stores based on consumer response Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 20, 2019
  • Karmakar, Moumita
    • Affiliation: College of Arts and Sciences, Department of Statistics and Operations Research
  • In this project, we determine base sale volume as a first step to identify effect of promotions on sale. Then we classify stores based on consumer response to promotions. We analyze three data sets which contain data on sale prices(in US dollars) in presence or absence of promotions. Dataset1: The first data set (containing 5 time series of two year daily sales transactions at different levels of product in absence of promotions) we perform sizer analysis on each of the raw data series to determine base sale volume. For each of five series mode represents the price (in US Dollar) around which maximum sales occur. We then take logarithm of each series and perform the sizer analysis for each. From sizer analysis of both raw and log-transformed series we can say that each of the five series is unimodal. Dataset2: For the second data set (containing 10 time series of two year daily sales transactions at different levels of product with many promotions), we do sizer analysis to assess the underlying pattern of the time series data as first step. To determine significant difference between promotion and non-promotion weeks, Distance-Weighted Discrimination (DWD) is performed on the second data set. From sizer analysis of raw data series we can say that except third and sixth series all series are unimodal and from log-transformed series except 2nd, 3rd, 4th and 6th all series are unimodal. DWD on second data set showed that for most of the the series differences between promotion weeks and non-promotion weeks are prominent and for rest of the series the difference are not that prominent. Dataset3: For the third data set (consisting of 300 weekly time series from 300 stores with many promotions), we perform various exploratory functional data analysis methods to identify difference between stores.From the FDA view plots and PCA projection plots for third data set, we can conclude that stores behave more or less the similar way.
Date of publication
Resource type
Rights statement
  • In Copyright
  • "... in partial fulfillment of the requirements for the degree of Master of Science in the Department of Statistics and Operation Research."
  • Marron, James Stephen
Place of publication
  • Chapel Hill, NC
  • Open access

This work has no parents.