2019 Facebook Data Challenge
Understanding the San Francisco Business Landscape
I worked on this project with a group of 3 other students at the 2019 Facebook Data Challenge, which brought together 40 students who were selected from around the country to compete in a 2-day datathon. The competition tested each team on their abilities to tackle a business problem by utilizing a data-driven business strategy, conducting statistical analyses, and preparing a presentation to Facebook data scientists and the Head of Facebook Data Engineering.
Overview
Tools Used: Python, R, Tableau, OpenRefine, Excel
Data Sources: SF Business Location Data (given for datathon), Demographics by Neighborhood (found on our own), and statistics on homelessness and food shortage in San Francisco
The questions we were posed were the following:
1. Provide an overview of the San Francisco business landscape.
2. What type of business should a person open?
3. Where should the business be opened?
4. What types of business licenses would be needed?
5. What are some applications of this dataset for Facebook?
Tools Used: Python, R, Tableau, OpenRefine, Excel
Data Sources: SF Business Location Data (given for datathon), Demographics by Neighborhood (found on our own), and statistics on homelessness and food shortage in San Francisco
The questions we were posed were the following:
1. Provide an overview of the San Francisco business landscape.
2. What type of business should a person open?
3. Where should the business be opened?
4. What types of business licenses would be needed?
5. What are some applications of this dataset for Facebook?
Methodology
We first cleaned the data.
This involved a combination of utilizing pandas and numpy in Python, some statistical packages in R, and OpenRefine and Excel for some basic cleaning. In order to get a big-picture overview of the data we were working with, we aggregated data to get information on the most common/least common types of businesses in the city, density of businesses in certain areas, and more.
Then we came up with a strategic framework to tackle the questions.
We knew that our end goal was starting a for-profit business that makes a positive social impact on the San Francisco community, so we started by determining the type of business to start. Our thought-process used a combination of qualitative knowledge that each of us had about the city along with quantitative findings to identify gaps. Once we figured out the problem we wanted to solve, we used our datasets to determine which area of the city had the greatest need.
Lastly we worked on the analysis.
The best type of business to open would be one related to food.
Our visualizations were created in Tableau, and a brief overview of our presentation can be found here.
We first cleaned the data.
This involved a combination of utilizing pandas and numpy in Python, some statistical packages in R, and OpenRefine and Excel for some basic cleaning. In order to get a big-picture overview of the data we were working with, we aggregated data to get information on the most common/least common types of businesses in the city, density of businesses in certain areas, and more.
Then we came up with a strategic framework to tackle the questions.
We knew that our end goal was starting a for-profit business that makes a positive social impact on the San Francisco community, so we started by determining the type of business to start. Our thought-process used a combination of qualitative knowledge that each of us had about the city along with quantitative findings to identify gaps. Once we figured out the problem we wanted to solve, we used our datasets to determine which area of the city had the greatest need.
Lastly we worked on the analysis.
The best type of business to open would be one related to food.
- Beyond meals provided by nonprofits and the government, the city's population is short 200 million meals per year
- Data showed that food businesses had lower rates of closure than other kinds of businesses
- Goal of business is to bring restaurant food waste to food pantries
- Used datasets to map out locations of food pantries
- Mission District has a large concentration of low-income residents, the target demographic for our business
- Mission District has the largest concentration of restaurants in San Francisco, making it a prime location for customers for our business
Our visualizations were created in Tableau, and a brief overview of our presentation can be found here.
Challenges
- Every team was made up of two students with a technical business background and two students with a heavy computer science background. This posed some challenges, as every team had to figure out how to balance the development of a business strategy with conducting enough heavy data analysis. My experiences working in this team taught me a lot about how to work with very technical people who may not be focused on the bigger picture, but also taught me how to look more deeply into specific, smaller problems that fit into the main problem at hand.
- Time, as in every datathon or hackathon, proved to be a huge issue. With everyone on the team being well-versed with different softwares, we had to settle on which programming language to use (we picked Python over R in most cases) and how to visualize data (we used Tableau). We ran out of time to ensure that the presentation of our data told a coherent, compelling story. This is something I hope to improve upon in future datathons.
