To: Amazon marketing and research departments

This report will focus on identifying how much each city spent on tech products throughout the months of 2019, finding trends, and providing information that might be helpful in improving sales from a marketing perspective.

The first thing that I wanted to do was see which city/zip spent the most money on Amazon during this time period. In order to do this I multiplied the ‘Quantity’ and ‘PriceEach’ variables (after making them numeric) to create a new variable called totalprice. I then summed every city’s totalprice’s to create this bar graph showing the total spend of each city: (I know these zip codes contain only a portion of these cities populations but I will be using ‘zip’ and ‘city’ interchangeably)

This gives us great insight but I was worried that it doesn’t show the whole story. Cities with a higher population will naturally have more purchases, so it is important to take that into account. I made this bar graph to show the cities populations:

It seems notable that the LA zip code has almost double the population of the others, yet still had significantly less sales than San Francisco. I wanted to see both variables together so I made this scatter plot that compares each city’s population with their total spending:

This gives a good idea each city’s relationship with spend and population, but I wanted to get a more exact idea of the ranking. To do this I made bar graph that compares the cities using a new variable I made called ‘spending per capita’ which divides each city’s total spending by their population to give us per capita values on their spending:

These visualizations show us that the zip codes in San Francisco, Seattle, and NYC spent the most per capita, and Austin, LA, and Atlanta spent the least per capita. In order to identify potential factors I want to compare spend per capita with other variables.

The first thing that I noticed is that the three highest spenders per capita had some of the highest median ages, I made this dot plot to explore the relationship:

As you can see, there is clearly a positive relationship between age and spend per capita in this dataset.

Next I wanted to analyze spending per capita against income statistics like income per capita and median household income:

There is a positive relationship between spend per capita and both of these income statistics. It only makes sense that people with higher income spent with more freedom.

The last thing I wanted to look at was if the diversity of these zip codes has any effect on spend per capita. I did this by dividing the white population by the total population, and then making an inverse of that variable to show diversity on a scale from 0-1. Here is diversity plotted with spend per capita:

As you can see, it seems that there is an inverse relationship. However, I noticed that LA and Austin seemed to be quite extreme outliers, and if you ignore them the relationship looked to actually be positive. I made the same plot but filtered out the two outlier cities:

So it is clear that LA and Austin were outliers, without them diversity has a more positive relationship with spend per capita than any of our previous variables.

I have found that the cities with an older, wealthier, and more diverse population were the best sellers in this dataset. The question of why the diverse populations of Austin and Los Angeles did not spend in these months is interesting to me, it is worth looking into how we can market to the populations in these zip codes. It may be that age is also a factor in these cities, as they are the second and third youngest populations respectively. It is clear that younger people were not buying in these months, which is another area we might invest in for the future.