This project was developed to get a general idea of Venture Capital funding throughout the world, to hopefully identify some key statistics or trends involving that funding. The data visualizations included in this web application showcase broad data aggregations across 3600+ cities on every continent (except Antartica) in an effort to find general funding trends.
The data was primarily sourced from Kaggle.com as a replication of Crunchbase data from the years prior to 2015. This data was supplemented by geocoding location data using the Google Maps Geocoding API. More details about the data are relayed in the following section.
This web application is available here, hosted on GitHub Pages as a data dashboard. Enjoy!
The primary source of our data was the investments_VC.csv file available on the Kaggle website here. This file includes all the startup names, total funding amounts, funding per round, start date, date of last funding, city, state, region, country, etc. that were used to create the data visualizations in this application. This data was cleaned from approximately 50,000 entries to about 37,000 entries by requiring valid, non-null values for the country_code, city, and total_funding columns. Additionally, the dates in the data set were limited to the years 2005 through the end of 2014, in order to get a more recent snapshot of VC funding throughout the world. As a side note, we did initially consider getting more recent data from Crunchbase.com, but downloading the data we wanted as a .csv file using their dashboard query tool required an "Enterprise" level account, which cost money and required a valid company and pre-defined application and use case for their data. This would be a good starting place for a more complete look at recent VC funding of startups in 2015 and after, but we could not afford that higher, "Enterprise" level of access for this basic project.
In addition to the data provided by Kaggle from Crunchbase, we needed to add a couple of columns for the geographical coordinates (Latitude and Longitude) of each city in the dataset. We aggregated the data by each city, so that the funding for each round and total funding represented the cumulative sum of all the funding for all startups in one city, and added a company_count column that represented the total number of distinct startups in a city. This reduced the data set to approximately 3800 cities spread throughout the world, and we then used the Google Maps Geocoding API to gather all the coordinate data (Latitude and Longitude) based on the city and country names. Aggregating the data by city in this way allowed us to plot all datapoints and analyze broad trends in the data relatively easily, without encountering major performance problems with our visualizations.
The type of database selected for this project was PostgreSQL, which is a common, performant SQL dialect that interfaces well with web applications and our data table. Due to time constraints, we did not break up the cleaned data into multiple tables with foreign keys, instead opting for a more monolithic table containing all information in the database. The SQL_Loader Jupyter notebook creates the "vc_funding_db" database within Postgres, and then defines, creates, and populates the "vc_funding" table within that database. As an installation note, you (the user) needs to include a config.py file in the same directory as that notebook, that contains two variables. There needs to be a user variable set to your Postgres server username, and a db_password variable set to your Postgres server password. These variables are different for every user, so make sure to include your own authentication credentials before trying to run the SQL_Loader notebook!
Please note that this config.py file will also be used when running the application, in order to connect to the database that holds the data for the application's APIs.
As a side note, the "vc_funding" table uses a composite primary key that is comprised of the (country_code, city) pairing of columns. This pair of columns is expected to be unique for all entries into the table, since the data is grouped by city, so each city and country should not have duplicate entries (it would have just been aggregated into one city in that country).
There are three visualizations in this dashboard application: a cluster map, a choropleth with world countries, and bar charts for VC funding rounds. The two geographical maps (the cluster map and choropleth map) were created with the Leaflet framework on top of a Mapbox map, and the bar charts were created using the Chart.js framework.
The cluster map for this web application is split into five different parts, which can be selected by clicking on the different "overlays" available from the menu on the top right corner of the map, which are shown as radio buttons.
The choices for these overlays are described as follows:
This value is initially selected when the cluster map loads. This layer shows circular markers that include the number of startups at a location, with the different marker colors depicting different sets of numbers. Green markers are used for values less than 10, yellow markers are used for values less than 100, and orange markers are used for values of 100 and greater. Zooming out causes the markers to group together with their neighbors, creating a cluster marker that represents a larger area on the map. Zooming in causes the markers to split apart into separate, smaller groupings of markers, enabling finer details of startup distribution in an area to be seen.
This layer shows markers with circular icons whose colors represent the order of magnitude of total funding at that location (in USD), with the actual amount of total funding as that marker location displayed in text (using scientific notation, since the values get rather large) to four decimal points. The legend for the marker colors is located in the bottom right corner of the map, and uses text as well as numbers to relay the order of magnitude, since looking at a large number with an extra zero or power is not descriptive enough for our purposes. For this layer, it is important to note the differences in marker color first, since that represents different orders of magnitude, and then to look at the number value after that in order to get a good understanding of the approximate VC funding into a geographical area.
This layer shows markers with circular icons whose colors represent the order of magnitude of average startup funding at that location (in USD), with the actual amount of average funding as that marker location displayed in text (using scientific notation, since the values get rather large) to four decimal points. The legend for the marker colors is located in the bottom right corner of the map, and uses text as well as numbers to relay the order of magnitude, since looking at a large number with an extra zero or power is not descriptive enough for our purposes. For this layer, it is important to note the differences in marker color first, since that represents different orders of magnitude, and then to look at the number value after that in order to get a good understanding of the approximate average VC funding per startup into a geographical area.
This layer shows markers with green circular icons whose areas represent the "market share" of that geographical location's startup count vs. the entire count of startups worldwide, with the precise percentage (to two decimal places) displayed below the marker. As the user zooms out, smaller icons combine into larger markers with higher percentages of the world startup count, and represent larger geographical areas. The reverse is true as the user zooms in on a location - the markers split apart into different locations and give more specific information about the percentage of startups in a smaller geographical area.
This layer shows markers with blue circular icons whose areas represent the "market share" of that geographical location's startup funding vs. the entire funding amount provided to startups worldwide, with the precise percentage (to two decimal places) displayed below the marker. As the user zooms out, smaller icons combine into larger markers with higher percentages of the world startup funding, and represent larger geographical areas. The reverse is true as the user zooms in on a location - the markers split apart into different locations and give more specific information about the percentage of startup funding in a smaller geographical area.
The choropleth map for this web application is split into three different parts, which can be selected by clicking on the different "overlays" available from the menu on the top right corner of the map, which are shown as radio buttons. Additionally, more details for each country are shown if the country is clicked on.
The choices for these overlays are described as follows:
This layer shows the company count per country on given a linear scale (see the legend in the lower right corner of the map for more details). The dominance of the US in startup count by country is apparent, with the US having over ten times more startups than the next country by startup count, which was the UK. As a result, the next choropleth was created to help visibly discern the second-tier countries by startup count.
This layer shows the company count per country (excluding the US) on given a linear scale (see the legend in the lower right corner of the map for more details). These countries can be viewed as second-tier countries for startups, to get a better understanding of notable countries with startups outside the US.
This layer shows the average funding per startup by country, with a detailed legend in the lower right hand corner giving the ranges of values that correspond to the shading levels. There are essentially three levels, or groupings of average startup funding (given by a k-means clustering algorithm), which represent the highest level of funding (or "lots" of funding), the middling level of funding (or "some" funding), and the lowest level of funding ("none"). The map can be used to visually show the geographic distribution of average funding per startup per country worldwide.
There are four different bar charts available on this data dashboard, which display the total and average funding amounts in USD ($) by round of funding for the top five US cities and the top five foreign cities, respectively. An example of the total funding bar chart for US cities is depicted below - the app itself has larger, better quality bar charts so only one bar chart example is provided here.
The first planned feature to be implemented is auto-loading the first cluster map when the app starts - this will engage the user immediately when the app is opened.
We are also planning on overhauling the interface by adding a "Help" or "User Guide" to the application, to enable new users to more easily navigate around the page and find information they are looking for. This will also include an explanation of the app's purpose, data used and served, and explanations of the map layers and bar charts.
Finally, if there is sufficient interest in the application, we could add a date filter to display data for ranges of years or particular years between 2005 and 2015.
This app was also designed to import new data automatically and could live-update using Crunchbase data, but their API subscription fee is cost-prohibitive at the time, so that feature is possible but not planned at the moment.