This project consisted of 2 practice clients to prepare us for real client projects. Alongside preparing reports for these clients, we gave in person presentations!
Client A
The first client, Booze ‘R’ Us, requested a method by which to predict future sales for growth purposes. We proposed a machine learning approach fitting a multiple linear regression model to historical monthly sales data from Booze ‘R’ Us’s storefronts and applied the modeling process to a case study similar in scope. The model was fit to historical data and accurately predicted future monthly sales for an average storefront. Furthermore, the features included reflected factors that we found tend to drive or have the greatest impact on monthly liquor sales. Through our analysis, we selected three main feature setups and evaluated each to finalize using month, size of bottles, price, and type of alcohol. In addition to developing a robust predictive model, we identified specific sizes, price ranges, and liquor types that serve as the primary drivers of sales.
Client B
The second client, Drinking Excess Alcohol is Dangerous (DEAD), was interested in the driving factors behind small and large alcohol purchases in Iowa. We proposed a machine learning model fitting a multiple linear regression model to a random sample of past sales data. The model fits sales data and predicts the quantity of alcohol purchased. The features of the final model represent the most influential factors driving alcohol purchasing behavior. This investigation has revealed valuable insights that can guide DEAD in shaping responsible alcohol retailing initiatives, with a focus on seasonal, category-specific, and college town-targeted campaigns, while upholding ethical principles of transparency and data privacy.