Amsterdam Wine Co. is a local winery that has been successful over the years because of its great wine selection. Our team believes we can further enhance Amsterdam Wine Co.’s wine selection by integrating analytics to the process of adding new items. It is crucial for businesses to start implementing these techniques to differentiate themselves from competitors and enhance the customer experience in an efficient way.
We analyzed three different datasets; The first dataset contained information of wines from different regions with its corresponding description, province, points, and price. The two following datasets (one for red wine and one for white wine) described the physicochemical qualities of different wines with its corresponding quality score. We cleaned, explored the data and performed different analyses in R to develop 3 recommendations.
Recommendation 1: Narrow down the wine options to the Top 50 wines based on our recommendation system.
Using a popularity-based recommendation system we constructed a list of the best 50 wines from the dataset that contained almost 120k wines. From these 50, the prices ranged from $25 to $1500 and the points ranged from 93- 97. With such a wide price range, we suggested categorizing the wines in three different price ranges (low-medium-high) and target different segments.
Recommendation 2: When choosing and selling wines, utilize the keywords from the word cloud (such as rich, black, age, structural, complex and balance).
Using text mining techniques, we created word clouds of the best wines (above 95 points) and the worst wines (below 82).
Top wines: rich, black, age, structural, complex and balance
Worst wines: tart, herbal, vegetable, green, bitter and flat
The word cloud for worst wines should be used to avoid selecting wines that have these keywords in their description. The one for the top wines should be used to choose new wines, as well as to describe the wines once added to the selection to appeal to the sensory palate of customers and accelerate the final purchasing decision.
Recommendation 3: Narrow down the wine options based on the established quality benchmarks.
We created one table for each wine (red and white) with the top 5 physicochemical features that influence quality and the averages by quality score. We suggested Amsterdam Wine Co. to follow these benchmarks and stay in the green range when selecting a wine. For example, a high quality red wine with a score of 8 has on average 12.1% alcohol, 0.768 sulphates (g/L), 0.995 density (g/mL), 0.423 volatile AC (g/L)and 0.0684 chlorides (g/L).
I would like to thank Alexandra Jehle and Rita Dai for collaborating with me in this project.
Below is the project's written report with a link to the folder that contains the R code.
댓글