Beer is one of the most widely consumed beverages across the world right after water, coffee, and tea. Beer consumption has kept increasing year after year. In 2020 only, despite the COVID-19 crisis that saw the whole world shutting down, people were not discouraged from downing an impressive 180 billions of litres. Beer always wins! One of the strengths of the beer lies in its diversity which largely results from the fermenting process. This diversity manifests itself in the existence of a plurality of beer styles, that can accommodate the complex desires of our discerning palate. Soft, fruity, strong, bitter, sour, there is something for everyone! Retailers have been poised to offer a smart selection of beers to increase customer satisfaction fostering their loyalty and, in the end, increase their sales. This is when we come into play.
A global retailer has appealed to us to help him redesign the beer shelves of their retail stores in various locations. This should be done in a smart way to help the retailer bolster its sales and differentiate itself from its competitors. Our idea is to help shops design their beer shelves to enhance shopper experience through personalized recommendations. Each shelf will be addressed to a different type of beer drinkers and contain a selection of beers styles among the ones favoured by drinkers of this category. Each beer style will also have a small textual description to further help the customer in his choice. The customer self-identifies to a category and is guided through a selection of beer styles briefly described that are popular among drinkers of his category. Additionally, since the preferences in terms of beers might be inherent to a country and vary across countries, the layout for the beer shelves should be adapted to the location of the store. Indeed, the selection of suggested beer styles depend on the preferences of the beer drinkers which might not be identical across countries.
The shelf layout that we will be suggesting to the retailer will be obtained through a data analysis of beer reviews. We have at our disposal two datasets from BeerAdvocate and RateBeer that consist of beer reviews. We chose to use the data from BeerAdvocate to carry out the following analysis. It spans a period of more than 20 years, including all ~8.4 million reviews up to August 2017. Each review includes ratings in terms of five "aspects": appearance, aroma, palate, taste, and overall impression. Reviews include product and user information, followed by each of these five ratings, and a plaintext review.