Katrin Gruber teaches how to build market response models using historical data to optimize marketing mix and predict customer behavior. She walks through a beer-sales dataset with 124 weekly observations, showing how to inspect data in R and calculate descriptive statistics before modeling the relationship between sales, price, advertising, and promotions.
Chapters
Welcome to my Data CAM course on response models for marketing research. My name is Katrin Gruber. I am an Assistant professor at the Department of Econometrics at Erasmus University Rotterdam. In this course you learn how to construct simple models of market response. The first part of this course starts with response models for sales price relations and how to extend this relation to to account for the effects of marketing activities and carryover effects. In the second part, we head over to response models for customer purchase decisions and to explain how customers react to marketing activities. To grow your business, you need a plan about how to communicate and sell your product to the customers. An effective marketing plan combines all the tools you are going to use to communicate the benefits of your product like advertising, public relations, word of mouth or price promotions. The key is crafting the right mix between these tools to achieve sales increases and market share goals. Market response models are widely accepted statistical tools used to optimize advertising mix and promotion tactics. Response models use past data to leverage information on the right marketing mix. Those information is later adopted by marketing professionals to adjust product prices and to optimize marketing strategies. Response models also enable to predict how customers might respond in the future and therefore allow to test the effectiveness of alternative marketing plans and to plan future marketing activities. In the practical example, we are working with a dataset named SalesData. The data contains weekly beer retail sales for a single store. In addition to sales, the data also provides information on prices and different advertising and promotion activities. Understanding your data is important to successfully build a meaningful response model. Therefore, we start with having a look at the data by using the function structure. The dataset contains 124 observations and six variables of different types. We obs which indexes the observation week. The numeric variables sales and price where sales is referring to the number of units of beer sold and price is referring to the average unit price within the respective week. Display coupon and display coupon are the advertising and promotion activities run by the company and are included as 01 coded integer valued variables to get a first impression about the overall performance in beer volume sales, calculating simple descriptive statistics like mean, minimum and maximum might be sufficient. This can easily be done by using the function mean, max and mean on the sales variable of the sales dataset. Per week, on average, around 119 units of beer are sold but with a huge variation as the minimum value is around 11 and the maximum value is around 1,400. Now it's time to.
