shambolic analytics

This exercise uses a kaggle dataset from an Ecuadorian-based grocery retailer, Corporación Favorita, accessed here. At this point, I already cleaned and pre-processed the data (converting datatypes, resolving nulls, merging csv files, etc.).

The purpose of this exercise was to answer questions with this dataset using both pandas queries & visualizations with matplotlib or the seaborn packages.

This was mostly done to practice with the seaborne package since I wanted to get faster at making graphs during my exploratory data analysis process.

There’s redundancy in my plotting code (repeated formatting and import calls) but this was intentional, as I wanted each code block to be a snapshot. That way, if I wanted to recreate the plot, I would be able to do so just by following along my notes here. If I were implementing this in a real use case, I would turn my plots into a function that had global formatting to reduce the repeated chunks of code.

Each question has an image of my code for the pandas query, a plot, and then an image of the code I used to make the plot. My reflections for this exercise are at the end of this post.

I used jupyter notebooks in VSCode for these exercises.

Question 1: which product families appear most often in the retail dataset? (Frequency)

Question 2: which product categories generated the most unit sales (top 10)?

Overall reflections

• plots and visualizations can be more effective at quickly communicating information
• the plots revealed more data cleaning was needed and helped me identify which ideas needed more queries and investigation
• plots take me more time to build and involve troubleshooting with layout & consulting seaborn documentation
• python is very efficient with queries for large datasets; running these in excel with macros would have taken a much longer time