INTRODUCTION
The Internet has given birth to some of the most powerful corporations ever. Google, Facebook, and other tech giants have quickly risen to fame thanks to its innovative business ideas and technologies. For better or worse, these companies are fundamentally changing the way people carry on their lives on a daily basis. One could say that a “parallel universe”, in terms of the digital, has been created with the commercialization of the Internet. Resultingly, people are living large fractions of their lives through smartphone and laptop screens.
One company that has taken advantage of this increasing screen time is Amazon.com. Currently ranked third among the highest valued companies in the world, Amazon gained its worth by conquering online shopping, widely known as e-commerce. Considering its immense size, it can be hard for people to comprehend just how large Amazon.com really is, including myself, as well as how the company actually makes its money, and so on.
This information visualization project will guide the observer through the company, and by providing visual presentations of various aspects of Amazon’s business, the project can be likened to a summarized visual company annual report. The project will not precisely follow the annual report but offer some of its most interesting parts, such as financial information, size in terms of employees, geographical facts, etc. The target audience of the end product is stakeholders or potential stakeholders who are, in some way, connected to the company. It might be someone who is looking to potentially purchase company shares, or an organization conducting research for some other purpose. Simply put, the project suits those who strive for high-level knowledge of Amazon.
METHOD
Data Collection & Manipulation
The data for this project was collected from a wide range of resources. Financial data was collected from S&P500’s data services platform Capital IQ, as well as from Yahoo! Finance. Opinion data was collected from Twitter, using the Python module twitterscraper. Country data was collected from the World Bank database. All other information was derived from the Amazon annual report of 2018.
Much of the data used in the project was not available in a downloadable table format, e.g. CSV. Hence, I manually created, organized, and transformed several data sets in Excel. Furthermore, much of the data had to be transformed from “wide” to “long” format, in order to reach the full potential when working in Tableau.
Software
As just mentioned above, Tableau was used as the main software tool for this project. This tool was selected mainly because it was a good fit for the purpose of the project, but also due to a personal desire to improve my skills and extend my experience of the software. In addition to Tableau, Excel was used for data manipulation, Python was used for data collection, and R was used for sentiment analysis to generate the opinion visualizations.
Dashboard 1: Amazon in Perspective
The introductory dashboard in this visualization project is an attempt to explain to the observer just how large Amazon is, which is done by putting Amazon’s monetary and employee size in comparison to small countries. The countries, Cyprus, Macao, Luxembourg and Iceland, were selected because they all have comparable populations to Amazon’s employee count and I believe that they are rather well-known despite their size. These countries also have a rather high GDP per capita, so the cumulative GDP of the countries could measure up against Amazon’s enterprise value. I thought about potentially adding larger, more well-known countries, however, by doing so I would have heavily distorted the graphs, as the other countries and Amazon’s values would become really small in comparison.
For monetary measurement, I used country GDP of the four selected countries to compare to Amazon’s enterprise value (market value adjusted for debt, cash, etc.) over a 10-year time period (2009 – 2019). I believe that this efficiently demonstrates the tremendous growth of Amazon over the last decade. I color coded the country graphs to correspond to the country flag color, and Amazon’s field was made orange. This orange color was kept to represent Amazon throughout the project, considering that it is the main brand color of the company. Annotation was added to highlight the most interesting point in the GDP vs enterprise value graph, i.e. the year when the enterprise value of Amazon surpassed the cumulative GDP of the selected countries.
Below the GDP graph, I added another comparison graph to present how the number of employees at Amazon relates to the population count of the four countries. For this visualization, I decided to conduct a different method to show the proportions in order to avoid any confusion. I created a bar chart of population count and shrank the chart bars to the minimum. On the dashboard, I added flag icons to represent the countries, as well as a circled Amazon logo. The icons were fitted to the bar height so that the height ratios would correctly represent the size proportions. This was done in accordance with the information coding heuristic which recommends using realistic characteristics to enhance understandability.
Dashboard 2: Financial Overview
For the financial overview dashboard, I decided to visualize the income statement, the balance sheet and sales breakdown in terms of product category as well as the breakdown of Amazon’s largest geographical markets. With four graphs, this dashboard page contains the most visualizations throughout the interface. I planned to add one more visualization, but as I felt that I was on the border to excessive detailing, which is a common mistake in dashboard design, I eventually decided to leave the fifth graph out. Particularly the financial dashboard is prone to excessive detailing, due to the complexity of financial graphs for observers with no experience in finance. The income statement was made into a waterfall chart, which is a typical visualization for sales and cost breakdown. The revenue build-up contains of sales category groups, which are also represented in a pie chart with matching colors. The pie chart is also used as a filter to highlight the amount of revenue from different sales categories. Since clicking on a pie chart to sort is not a traditional way of applying a filter, I added a simple instruction text next to the chart in order to signify its functionality.
I decided to go with contrasting colors for the balance sheet, to differentiate the asset and financing sides. To enhance the understandability of the balance sheet, I added icons to signify the different parts, as well as percentage labels.
Finally, the geographical market breakdown was made into an area chart, and the areas were colored in a brown/orange scale in order to avoid overuse of color and keep the coloring consistent. The brown/orange scale was selected because it looked nice together with the overall dashboard color scheme. For discoverability, I added map icons to function as signifiers for the chart areas, along with a short text description containing the percentage share of each market.
Dashboard 3: Locations of Operation
This dashboard presents Amazon’s locations of operation in the US in the form of an interactive thematic map. I thought this would be interesting to show, as you otherwise mostly think of Amazon as a company operating in the “cloud”. With this visualization, we are able to observe their physical presence.
I tried several combinations of various base map layouts and different data point colors. Eventually, I decided to stick with a dark mode map and colorful data points. I figured that this design fits well into the overall aesthetics of the project, and it looked rather “techy”, in accordance with Amazon being a tech company. In addition to pure aesthetic design choices, I decided to remove state names, country names, landscapes, and more, to keep the map as clean as possible and maintain focus on the data points. The state names would appear when clicking on a data point, which I believe was enough.
I assigned four different colors to the data points; orange for headquarters, red for software development, blue for store location, and green for warehousing and fulfillment. There was no particular reasoning for choosing which color represented which entity, except for headquarters being orange as this was most heavily associated with Amazon.
Next to the map, I added a ranked bar chart, showing which states have the most locations for each entity. Above this chart, the user will find functions to filter the data accordingly.
Dashboard 4: Amazon Perception – Ethics & Opinions
I found a good Wikipedia article that gave an overview and explanation to all Amazon criticism, which I decided to incorporate in my presentation. The site was attached as a website object in the Tableau interface, allowing the user to quickly look for Amazon criticism. My only concern for the Wikipedia attachment was that there are no buttons to move back and forth across webpages, so if observers were to accidentally click on a link, the only way to get back to the relevant Amazon article is to right-click on the page and go back that way.
In addition to the website, I conducted a sentiment analysis on Amazon by scraping Twitter and analyzing the tweets using Python and R. The sentiment analysis was transformed into a temporal area chart, showing the proportion of negativity and positivity expressed toward the company between January 2018 and December 2019. The data was collected by applying regular expression for finding tweets containing #AMZN, causing the gathered tweets to be focused on the stock market. Hence, a slightly transparent share price time series made sense to add to the sentiment time series. Overlapping time series data in this way allows the user to investigate potential correlating patterns (Few, 2009).
I also added a bar chart to the dashboard, which presents the feelings that were most associated with Amazon over the analyzed period. This provides a more granular understanding of the public perception of the company. I color-coded the feelings to correspond to colors that I assessed to be positive and negative. Red, orange, grey, brown, and green/yellow were used for negative feelings, while positive feelings were assigned green, blue, and pink colors. The meanings of color vary over different cultures, but I went with my personal opinion. There is also a risk of getting the colors mixed up due to color blindness among some people. However, color choices for this bar chart were not crucial. The graph would still work in monochrome mode, as the bars are described with text as well. Hence, I kept my initial color design.
UX Research Method
I recruited three people to participate in a user test of the Tableau visualization interface. They all shared the common interest of potentially purchasing shares in Amazon. The test was designed as a combination of task completion and questionnaire (see attachment above). First, the participants observed the interface freely to become familiarized, then some questions about the information were asked. After this, the participants gave their opinion on four statements, e.g. how easily they found the answers for the questions asked. Finally, they were asked to provide some qualitative feedback about the visualizations. I designed the UX research method in accordance with usability theory methods, to ensure a qualified end result. For example, a scenario was explained prior to the test in order to get the users into the right mindset.
General Design Rationale
The size of the dashboards was set to 1200 x 800, based on laptop screens. Hence, a laptop is the recommended hardware for viewing the project, although larger and smaller screens will also work. For smaller screens, the user will have to move around over the page in order to view all the content. The information provided by the visualizations is rather high-level, making it more fit for the executive-level audience. In this case, however, as stated previously, the target audience is potential individual investors.
Going into the project, I created a custom color palette by altering the code in the Preferences file in the Tableau Repository. Hex codes corresponding to the Amazon brand colors were added to this palette which I used throughout the dashboard visualizations. All four dashboards were assigned an orange background color, based on the Amazon logo color. The logo itself was added to the top left corner of every dashboard to further add to the project branding. Buttons were added to the top right of the dashboards to facilitate navigation between the different layouts and to enhance the storytelling by making the user follow a patch, i.e. forward or back. The final commonality among all dashboards is that they were all assigned a title for describing the content of each dashboard. Again, the title was colored in orange in harmony with Amazon branding. Filtering options were added to the financial and location dashboards, other than that, the dashboards were kept rather free from user actions in accordance with the minimal action heuristic.
FINDINGS
Design
Regarding design and storytelling, I believe that the major finding for me in this project was what a difference it makes to combine color use with realistic characteristics. It adds a lot to the visualizations when color is not overused, but instead, the designer takes advantage of icons, etc. to add variety and understandability to visualizations.
I also felt that dashboards become a lot better when combining different types of charts and graphs, given that they are used with meaningful purpose and not only for the sake of variations. I particularly felt that the financial overview dashboard turned out well, despite being the most information-heavy page. I believe that this was largely thanks to a solid variety of graphs and colors.
Amazon
The combination of dashboards made a good visual summary of Amazon.com, and I felt that there are several interesting findings to be made by studying the graphs and charts, particularly if you are not previously familiar with the company. First off, I think it was interesting to see just how large Amazon is, as the combined GDP of four countries, although being rather small, became tiny in comparison to enterprise value from 2015 onwards. I also believe that you get a good idea of just how large the company is in terms of employees, almost being able to populate two Iceland.
From the financial information, I thought it was interesting to see how retail products corresponds to not more than 62% of the sales, as Amazon is known mainly as an e-commerce giant. Their cloud services make up for a whopping 13%, and I would assume this is a growing category.
I also found it interesting to be able to see how Amazon is perceived by the public on Twitter. I believe that it was interesting to see how the top three feelings expressed on Twitter were very positive and that the feeling ratio was predominantly positive over the last two years. I feel like a giant corporation like Amazon would naturally be a target of a lot of negativity, especially on Twitter, which proves not to be the case.
From UX Research
The user testing of the interface was successful, as I received feedback for which visualizations needed to be redesigned in order to enhance the understandability of the interface. All of the participants agreed that the income statement waterfall chart was the most difficult to comprehend. Although the participants were able to understand the information after a while, they provided me with recommendations on how the chart could be improved. As a result of this, I added the words “profit” and “cost” to each side of the chart and color-coded the words to convey with the bar colors. See visualization before and after UX-test feedback below:
In addition to the waterfall chart, I redesigned the enterprise value vs GDP graph, as there was an issue with the color legend box disappearing behind the graph. I moved the location of color legend to the side of the graph, and I added a text box with additional information. Not only did this redesign solve the issue above, but the location of the legend became more logical, sitting closer to the actual graph areas. See below:
Except for the two redesigns presented above, the participants were satisfied with the interface and enjoyed the information provided, as well as the aesthetics.
DIRECTIONS FOR FUTURE PROJECTS
The next steps for this interface dashboard would be to develop with additional parts, such as competitors, partners, subsidiaries, and more focused criticism and opinion information, in order to make the project appear more of an exact visual summary of Amazon’s annual report. Although these additions could be made, I still think that the most important and interesting parts were covered in this particular project, in regards to gaining a high-level understanding of the company.
It would also be interesting to perform a similar visualization dashboard on all the top tech giants (FAANG), in order to compare these corporations with each other.
REFERENCE(S)
Few, S. (2009) Now You See It. Oakland, CA: Analytics Press.