The goal of this project was to find the best spot/city (worldwide) to place the offices of a newly created company in the gaming industry taking into account the preferences of its employees.
The starting point is a collection composed of roughly 18,000 companies, which needed to be filtered according to the needs of the company, mainly:
- To be in a city with a gaming industry hub: this is a gaming company, which would like to outsource and establish collaborations with other companies in the industry. Therefore, being surrounded by other gaming companies is a must.
- To be in a city with successful (>$1M money raised) companies in the design and tech industries, as it is required by roughly 40% of the employees (designers and developers).
Thus, filtering the collection using these two initial requirements would help us identify those cities with a potential hub.
See below the filter applied to select gaming companies:
# Necessary filters: Companies whose 'offices' field is not empty and who have raised more than 1M $/€
office_filter = {'offices' : {"$ne" : []}}
money_filter = {'total_money_raised' : {"$regex" : ".*\d+[MB]"}}
necessary = {"$and" : [office_filter, money_filter]}
# If I only consider gaming companies:
tag_filter = {'tag_list' : {"$regex": ".*gaming.*|.*game.*"}}
overview_filter = {'overview' : {"$regex": ".*gaming.*|.*game.*"}}
projection = {"_id":0, "name":1, 'total_money_raised':1, 'offices':1, 'tag_list':1}
companies = list(c.find({"$or": [{"$and": [necessary, tag_filter]},
{"$and": [necessary, overview_filter]}
]}, projection))Out of the ~18,000 companies, 285 are purely gaming companies (gaming hub). If design and tech tags are included in the filters (gaming-tech-design hub), then 430 companies remain. The top five cities in each case are:
Note: less nº of companies are observed in San Francisco when considering the gaming-tech-design hub because an additional categorical filter was applied in order to narrow the search.
Based on this, four cities were chosen for further analysis: San Francisco, New York, South San Francisco (for its proximity to San Francisco) and London.
- First, the filtered companies were mapped on each of these cities.
- Then, the most centric company (the one surrounded by all the other companies) was chosen to minimize the distance to all the other companies and its coordinates were used for requesting the remaining requirements of the employees in Foursquare.
The following figure shows the maps of San Francisco (left) and New York (right), and the markers correspond to the filtered companies. The red arrow indicates the most centric company.
Based on the employees' requirements, foursquare was used to request information. The function was built taking the following parameters:
- The query in foursquare (string).
- The city (corresponding to the coordinates of the most centric company).
- The radius of the search (first number, in meters).
- The category in foursquare (second number, a foursquare ID).
- The maximum number of results to return (limit).
The resuls were sorted by distance. The requirements of the employees can be seen below in an extracted piece of code:
schools = getFoursquareCategory("school", city, 2000, 12058, token_fsq, limit=5)
clubs = getFoursquareCategory("night club", city, 2000, 10032, token_fsq, limit=5)
starbucks = getFoursquare("starbucks", city, 2000, token_fsq, limit=5)
basketball = getFoursquare("basketball stadium", city, 10000, token_fsq, limit=5)
dog = getFoursquareCategory("dog grooming", city, 10000, 11134, token_fsq, limit=5)
airport = getFoursquareCategory("airport", city, 50000, 19040, token_fsq, limit=5)
vegan = getFoursquareCategory("vegan restaurant", city, 2000, 13377, token_fsq, limit=5)- Next, the distance between the chosen coordinates and each category (query in foursquare) was extracted from the results of the request.
- In the case of the nearby companies, which were also included in the final table, the distance was calculated using a function that considered their latitude and longitude coordinate values. Note: only the companies within a 5 km radius were considered.
- Each category was assigned a weight value according to their importance. This is a value that goes from 0 to 1, and adds up to 1 all categories considered. The importance was decided based on two main factors:
- The amount of employees that have that requirement.
- How common it is to find that requirement in a city.
- For each category, an average distance was calculated. This favoured having more options where to choose within a given category. The average distance was then multiplied by the respective weighting factor.
- If a requirement was not found for a given search radius (NaN in 'counts' and 'sum_distance'), it was penalized by adding that radius as the average distance (see the table below from South San Francisco as an example):
In order to calculate a final score, the weighted averaged distances were added. The final score was normalized as percentage based on the minimum value of the comparison. Thus, a score of 100% is the perfect score between the compared cities, and values (N) above 100 indicate that the categories are, on average, + N % more distant from the company's coordinates.
And the winner is... San Francisco! The coordinates of the potential company are: (37.781929, -122.404176).
- New York came as a close second. Interestingly enough, New York lost to SF due to the distance of the company to other gaming-tech-design companies, which are more scattered than in SF. On the contrary, the other activities/categories are more packed around the company's coordinates than in SF (see maps below).
Future improvements:
- Instead of analyzing one company (location) per city, iterate over all the gaming companies in SF and New York in Foursquare. Then, calculate the weighted average distance of each category/activity for each company in order to obtain the best location within the city.
- Complement this information with real state data. Is the difference in pricing (renting a property) of distinct neighbourhoods big enough to justify one location over another one?
Finally, interactive maps were built for each city. Markers were added and visualized with different icons and colors depending on the category. For clarification, see the legend below the pictures.
Click here to see the maps of: San Francisco, New York, South San Francisco and London.
Legend:
| Category | Icon | Color |
|---|---|---|
| The location of the company | Diamond | Black |
| Airports | Plane | Purple |
| Basketball stadiums | Football ball | Orange |
| Night clubs | Cocktail glass | Red |
| Dog grooming | Paw | Lightgray |
| Schools | Graduation cap | Blue |
| Starbucks | Coffee mug | Gray |
| Vegan restaurants | Cutlery | Green |
| Gaming / tech / design companies | Building | Lightred |