Promo and Calendar Effects on Grocery Store Sales Volume, A Shiny App
The skills the authors demonstrated here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
The project for this blog post is to develop an interactive Shiny App to Analyze Grocery Store Sales from a data collection of Grocery Stores owned by La Favorita in Ecuador.
Who is the intended user?
Various Decision makers in the La Favorita corporate system. Such as store managers who can use the dashboard to evaluate sales of products for their specific stores. Also, Operations Analysts, Buyers, Stock Managers to determine patterns between store clusters or geographic areas.
In this post I will discuss use cases and actionable insights from accessible through the dashboard.
Specific Goals for this Dashboard
A fully fleshed out app for analyzing sales would include many analysis tools. For this Shiny App I address a few key analysis questions:
1. How is Sales Volume for product Categories affected by promotional discount prices?
2. How Does Sales Volume vary over multiple time scales: day of week, month, year?
Some quick background notes on the data
Data Details - the dataset includes several items:
- Item Categories - supplemental information about item categories
- Daily historical data from January 01, 2013 to August 15, 2017
- Stores - supplemental information about the shops (city, state, etc)
- Transactions - the daily total sales transactions for each store
Wages in the public sector - are paid every two weeks on the 15th and on the last day of the month.
General Product Family Insights Based on Data
looking at the following screenshots as illustrative examples which represent several general patterns.
- Saturdays and Sundays are consistently the biggest shopping days of the week
- There is a slight uptick in sales after public sector paydays (the 15th and last day of the month)
Specific Product Takeaways From Data
After making use of the dashboard to spot specific category patterns. Alcoholic Beverages Drop Heavily after Saturdays. This decrease can be correctly can be quantified with follow up analysis.
Now, if we select two product categories we can compare them side by side. For example, with Dairy and Bread/Bakery both selected we can see some clear differences in long-term sales volume. Obviously, there is a dramatic increase in the sales of dairy around the final quarter of 2013 onward. Meanwhile, Bread/Bakery have a gradual but steady growth over time
These are a small sample of individual product insights. The advantage of a shiny app is the more you the more you can find.
Staples, like eggs, are less affected by promos than other categories. Looking at the egg promo analysis page we can see this pattern because in the box plot the mean sales volume on promo days in blue is quite close to the mean sales for non-promo days in light red.
Next we can see that Friday Promos affect some categories, Poultry and Seafood, significantly. Another pattern that occurs in several product categories can also be seen in the following screenshot, and it is that the impact of a promo seemingly depends on the day of week the promo takes place.
Final Comments and Practical Applications
Patterns in promotional effects can be used to clear inventory prior to expiration, or compensate for lower than predicted sales. Using promos on the days where they are most effective but not other days could allow for more profit on low-impact days and more on high-impact days.
The opposite effect can be used to prevent stocking shortages by keeping items off promotional discounts.
The overall goal of a dashboards like this is to spot initial trends and identify patterns that can be used directly or inspire additional in-depth analysis.