Recommendation on opening a New Restaurant - Using Data Science

A.1 Introduction

The age of intuition is undergoing a shift in today's world. Even though it will not totally go away, the art of decision making through that 'Gut-Feeling' is slowly being augmented (if not replaced) by the decision making through Data-points.

There are innumerable source of data. Some data are available free and some are available in lieu of payment. Regardless, data is available on any subject; be it data about climate for the last 30 years, data about automobile emission for different makes or data about consumption pattern of consumers sliced and diced in N-number of dimensions. In fact, the new problem could well be to find the way around the information available with us.

This article/write-up is a very simplified attempt to find a solution to a business problem using available data.

A.2. Description of the Problem Statement

Let us assume that a friend of mine has expressed his desire to open a new Restaurant. He wants to avoid large cities like New York and San Francisco. Now, as a friend and consultant I have to give him a recommendation.

Let us see how we can approach this problem in a step-by-step manner taking the help of "data". First let us see what all data-source can be taken to move towards the conclusion.

B.1 Source(s) of Data

Data had to be collected from multiple sources and stitched together to do the analysis and recommendation. 

First, I searched the Internet to get a listing of the cities and their population. This was available in Wikipedia link mentioned below.  

https://en.wikipedia.org/wiki/List_of_United_States_cities_by_population  - The census data was from 2018 but that was a close match. 

Foursquare API was used to gather the information about existing venues in the two cities. 

C.1 Methodology

The Wikipedia site lists the 314 incorporated places in the United States with a population of at least 100,000 as on July 1, 2018, as estimated by the United States Census Bureau. Five states—Delaware, Maine, Vermont, West Virginia and Wyoming—have no cities with populations of 100,000 or more. 

The technique of screen-scrapping was used read the city information into a panda DataFrame. It was found that two cities namely Chicago and Houston had population in the range of 2.5million. These two cities were picked up as possible choices for exploring the potential of opening a new restaurant. As of 2018 estimate, Chicago had a population of 2.7 Million while Houston had a population of 2.3 Million. 

Population constitute the most important factor for a Restaurant. There are other factors such as City culture, Per-capital income, ethnicity and even climate. But the remaining are secondary.  

The second data source was the Four-square API. Using the Explore option, the venues were read into panda DataFrames. While invoking the ‘Explore’ API for both the cities, a radius of 750 meters were taken and the category of “food” was taken. This filtered out other venues like Museums, theatres etc. So now we had the list of food joints of the two cities along with the categories. We got a total list of 588 food venues for the two cities 

Next, from the list of the food venues, the categories of Bistro, cafes, joints, food-truck were filtered out. So now we had a clean list of pure-play restaurants in the two cities. 

The next step was to generate a pivot table to give a summary of the restaurants type for the two cities in a format for easy comparison. 

D.1 Results

From the above tabulation, the observations are as below.

a.     It appears that Mexican Restaurants are quite numerous in both the locations (6 and 5). Opening a Mexican Restaurant in either city is going to face competition

b.     There are 5 Italian restaurants in Houston but only 1 in Chicago. There is a potential business opportunity

c.     Chicago does not seem to have a steakhouse at all. That should be a business opportunity here.

E. Observation & Recommendations

Overall, recommendation should be to go for a Italian Restaurant or a steakhouse in Chicago.

Now, as far as the location is concerned, a map-plotting using Folium shows that the restaurants are concentrated on two particular stretches of the city. In other words, they for two clusters. It is advisable that the new restaurant be opened in one of the clusters. It might seemingly be counter-productive but there is a rationale to this. Citizens of Chicago would have a perception or mind-share about the two clusters even if they are not formally identified/indicated. On a typical holiday or weekends, people wanting to dine-out would head for these clusters keeping the option open for a type of cuisine/restaurant. Having the new outlet there would have a potential footfall from the crowd thronging to those areas.

F.1 Conclusion

It needs to be mentioned however, that there are two factors which should have been included in the feasibility analysis. These are i) Per-capita income and ii) Ethnicity of the population(% distribution and not just the names). These data points were not readily available and hence these factors could not be included in the analysis.

Needless to say, higher the per-capita income , more the money at disposal and better the chance of a restaurant to attract patrons.

Ethnicity also drives the foot-fall. While nothing as such prevent a person of Greek origin to visit a Russian restaurant, it is generally observed that a person living away from his /her origin will have a sentimental attachment to culture and cuisine that he left behind. The sentimental attachment will drive him to the restaurant tied with his origin.

Hopefully, future research in this aspect will include the two factors mentioned.

The End