Marketing Analytics Project - Spring, 2017

Rafael Guimaraes, Brian Kangogo, and Nicholas Matthew

Date April 27 2017

Overview of The Starbucks Corporation and Our Project

Founded in Seattle, WA in 1971 by two teachers and a writer, Starbucks Corporation is the world’s largest coffeehouse chain with 23,768 locations worldwide, 13,107 of which located in the United States as of 2016 [1]. The company took a strategic turn in 1981, when Howard Schultz, the corporation’s current CEO, walked into Starbucks and envisioned the company modeling itself after the Italian coffee houses he had seen in Europe: a retail location that not only sells excellent coffee, but a company that sells an experience – a “home away from home” [2].

In 2011, the owner of Cafe on the Common, a small coffee shop on Main St. in Waltham, MA confided that he had attempted to recruit Starbucks to flip his store into a franchize. For reasons that weren’t fully clear, Starbucks turned down his offer saying Waltham was outside of their targeted market. Today, Waltham now has two Starbucks locations that didn’t exist in 2011: Brandeis University and 12 Market Place, both of which are accessible by higher-income individuals.

According to our research, 83% of Starbucks locations cater to white communities. Recently, the organization has attempted to bridge the gap with low-income and minority populations. In October, 2016, the company announced that it would be opening 15 stores in low-income communities of color [3]. Starbucks has also joined a coalition of businesses that have pledged to hire 100,000 at risk youth as a CSR campaign [4], and launched a racially-focused CSR marketing campaign called #RaceTogether, where the company encouraged dialogue about race relations [5]. The campaign didn’t fair well, with signs near cash-registers posing uncomfortable Madlib-esque conversations starters like, “In the past year, I have been to the home of someone of a different race ___ times” [6].

Objectives of Our Data Analysis

The guiding question for this project is, “How might customers’ average household-income inform Starbucks decisions to enter prospective markets?” We hypothesized that, yes, there would be a positive correlation between a community’s mean household income and the number of Starbucks locations, but that there might be an especially strong negative correlation with not low-income, but middle-income communities. We reasoned that middle-income, working-class populations would be more likely to exhibit financial discretion.

This R Markdown will first:

  1. Review data analysis of Starbucks locations found on Kaggle.com
  2. Present novel findings about a) Starbucks ownership and b) locations in the United States and Massachusetts as it pertains to average household income.

Previously Existing Kaggle Kernels


Map #1: Interactive Map of Starbucks Locations Worldwide.


countries <- df %>% group_by(Country) %>% summarise(Total = round(n()))
names(countries) <- c("country.code", "total")
countries$iso3 <- countrycode_data[match(countries$country.code, countrycode_data$iso2c), "iso3c"]

data(worldgeojson, package = "highcharter")
dshmstops <- data.frame(q = c(0, exp(1:5)/exp(5)),
                        c = substring(viridis(5 + 1, option = "D"), 0, 7)) %>%  list_parse2()

highchart() %>% 
  hc_add_series_map(worldgeojson, countries, value = "total", joinBy = "iso3") %>% 
  hc_colorAxis(stops = dshmstops) %>% 
  hc_legend(enabled = TRUE) %>% 
  hc_add_theme(hc_theme_google()) %>% 
  hc_mapNavigation(enabled = TRUE) %>%
  hc_title(text = "") %>%
  hc_credits(enabled = TRUE, text = "Sources: Starbucks Store Locator data by Github user chrismeller", style = list(fontSize = "10px"))


Notes:

  • Perhaps interesting to note that although Starbucks modeled itself in the 1980’s after Italian coffee shops, there are zero located in Italy.
  • Low income countries, namely those in Africa, Greenland/Iceland, also lack Starbucks.
  • China has a surprisingly high number of locations, higher even than Canada.
  • It is interesting that Starbucks is not in some high income African countries like Nigeria and Kenya. These countries have coffee shops that sells coffee even more expensive than starbucks e.g Java in Kenya. This shows that income is not the only factor that determine the location of a store
  • For future analysis, it might be interesting comparing the number of Starbucks per-capita for not only countries, but cities. Asia might have a higher number than those in Europe, despite being culturally distinct from western countries, raising the question as to why that might sociologically be the case.




Graph #1: Top 10 Countries Ranked by Number of Starbucks.

countries <- df %>% group_by(Country) %>% summarise(Total = n(), Percentage = round(Total/dim(df)[1] * 100, 2)) %>% arrange(desc(Total)) %>% head(10)

hchart(countries, "bar", hcaes(Country, Total, color = Country)) %>% 
  hc_add_theme(hc_theme_google()) %>%
  hc_title(text = "")


Notes:

  • It’s impressive to see how the United States has nearly five times the number of Starbucks than the next nearest country.
  • It is interesting that one of the fastest growing and second most populous country, India, does not make it on the top ten markets.
  • Again, it would be interesting to see how many Starbucks there are per capita.


Graph #2: Top 10 States with Starbucks

stores_usa <- filter(df, Country == 'US')

States <- stores_usa %>% group_by(state) %>% summarise(Total = n(), Percentage = round(Total/dim(df)[1] * 100, 2)) %>% arrange(desc(Total)) %>% head(10)

hchart(States, "bar", hcaes(state, Total, color = `state`)) %>% 
  hc_add_theme(hc_theme_google()) %>%
  hc_title(text = "")


Notes:

  • It might be interesting to attempt to tease out variables of why California and Texas are both strong markets for Starbucks.
  • It is interesting to note that the list of top 10 starbucks location almost mirror the ranking of US states by population. Six out of 10 are also top 10 countries by population


Graph #3: Starbucks Ownership Type


ownership <- df %>% group_by(`Ownership Type`) %>% summarise(Total = n(), Percentage = round(Total/dim(df)[1] * 100, 2)) %>% arrange(desc(Total)) %>% head(10)

hchart(ownership, "bar", hcaes(`Ownership Type`, Total, color = `Ownership Type`)) %>% 
  hc_add_theme(hc_theme_google()) %>%
  hc_title(text = "")

Notes:

Unlike Mcdonalds that uses a franchising model, Starbucks uses licensing model to expand.

  • Licensed: This is a form of an agreement where one company grants another party permission to use its patents, trademarks, copyrights and trade secrets.
  • Franchise: Similar to licensing but in addition to using intellectual property, the franchisee also has access to the operating system. This include franchisor’s distribution system and marketing campaigns.


Novel Data Analysis and Representation


Map #2: Starbucks Locations Point-Plotted on a World Map


world <- map_data("world") #Generates a world data by longitude and latitude


#plot All states with ggplot
world_map <- ggplot()
world_map <- world_map + geom_polygon( data=world, aes(x=long, y=lat, group = group),colour="white", fill="grey10")
world_map <- world_map +  geom_point(data=df, 
                               aes(x=Longitude, y=Latitude), colour="yellow", 
                               fill="red",pch=21, size=.5, alpha=I(0.7))
world_map


Notes:

  • Benefits of this graph allow the user to better see outlier locations, such as the Starbucks in South Africa, Kazahkstan, and the east coast of Australia.


Map #3: Starbucks Locations Point-Plotted on a Map of the United States.

all_states <- map_data("state")


#plot All states with ggplot
US <- US[US$state!="AK" & US$state!="HI" & US$state!="PI" & US$state!="VI"
         & US$state!="PW" & US$state!="MP", ] # Removing the non-continental states


US_map <- ggplot()
US_map <- US_map + geom_polygon(data=all_states, aes(x=long, y=lat, group = group),colour="white", fill="grey10")
US_map <- US_map +  geom_point(data=US, 
                              aes(x=Longitude, y=Latitude), colour="yellow", 
                              fill="red",pch=21, size=.5, alpha=I(0.7))

US_map


Notes:

  • As expected, majority of starbucks locations are in East and West coast. It is also interesting to see a lot of starbucks stores in Colorado.


Map #4: Starbucks Locations in Massachusetts.

MA_state <- map_data("county")

states <- subset(MA_state, region %in% "massachusetts")

MA_map <- ggplot()
MA_map <- MA_map + geom_polygon( data=states, aes(x=long, y=lat, group = group),colour="white", fill="grey10" ) 
MA_map <- MA_map +  geom_point(data=MA, 
                     aes(x=Longitude, y=Latitude), colour="yellow", 
                     fill="red",pch=21, size=1, alpha=I(0.7))

MA_map


Notes:

  • We were surprised to see the number of starbucks locations on Martha’s Vineyard and Nantucket: 0. We expected to see more, given the excellent summer market opportunity for seasonal revenues. On the other hand, these locations might have city-governments that have policies against corporate chains.
  • Poor Franklin County in north-western Massachusetts lacks locations.



Graph 4: Starbucks Ownership Type by Country


x<- df %>% select(Country, `Ownership Type` ) %>% group_by(Country,`Ownership Type`) %>%
  summarise(sum_Country=n()) %>% arrange(desc(sum_Country),desc(`Ownership Type`))


Country_own <- filter(x,Country=="CN"|Country=="US"|Country=="CA"|Country=="JP"|
                        Country=="KR"|Country=="GB"|Country=="MX"
                      |Country=="TW"|Country=="TR"|Country=="PH")

#Graph showing Starbucks Stores By country and ownership Type
ggplot(data=Country_own, aes(x=Country, y=sum_Country, fill=`Ownership Type`))+
  ylab("Number of Starbucks per OwnershipType")+
  geom_bar(stat = "identity")

Notes:

  • Interestingly, all of the franchises are in Great Britian.
  • It is very interesting to see that a lot of joint ventures are in Asia. This is a strategic move to penetrate this market. Most multinational corporation find it hard to make it in some of these countries. For example, Uber was forced to sell majority shares to Didi in China because of intense competition.

Graph 5: Starbucks Locations by City


by_city <- df %>% group_by(City) %>% summarise(Total = n(), Percentage = round(Total/dim(df)[1] * 100, 2)) %>% arrange(desc(Total)) %>% head(10)

hchart(by_city, "bar", hcaes(City, Total, color = City)) %>% 
  hc_add_theme(hc_theme_google()) %>%
  hc_title(text = "")

Notes:

  • Despite the US having five times more Starbucks locations than other countires, the top three cities are not in the United States (1. Shanghai, 2. Seoul, and 3. Beijing). .



Map #5: Heat map of Mean Household Income by State with Starbucks Locations.

## Parsed with column specification:
## cols(
##   fipscode = col_character(),
##   postalcode = col_character(),
##   Name = col_character(),
##   poverty_est = col_character(),
##   percent_poverty = col_character(),
##   household_income = col_character()
## )
heatmap <- ggplot(heat, aes(long, lat))
heatmap <- heatmap + geom_polygon(aes(group = group, fill = income),colour="white") + scale_fill_continuous(low="white", high="#006600", limits=c(30000,76100))
heatmap <- heatmap +  geom_point(data=US, 
                               aes(x=Longitude, y=Latitude), colour="yellow", 
                               fill="red",pch=21, size=1, alpha=I(0.4))
heatmap

Notes:

  • As expected, starbucks locations are located in high income states both in the coastal cities and midwest.
  • The map above exhibits rendering glitches in Calfornia and Colorado - artifacts of missing mean household income for those locations within the states.

Map #6: Heat map of Mean Household Income by County in Massachusetts with Starbucks Locations.

heat_map_MA <- ggplot()
heat_map_MA <- heat_map_MA + geom_polygon(data=heat_ma, aes(x=long, y=lat, group = subregion, fill = income),colour="white") + scale_fill_continuous(low="white", high="#006600", limits=c(30000,95000))
heat_map_MA <- heat_map_MA +  geom_point(data=MA, 
                               aes(x=Longitude, y=Latitude), colour="yellow", 
                               fill="red",pch=21, size=1, alpha=I(0.7))
heat_map_MA

Notes:

  • A finer map of Massachusetts confirms that Starbucks are located in high income locations
  • Boston has by far the highest number of Starbucks. This can be attribute to both high income and population
  • The lower income counties of Berkshire and Hampden have lower number of starbucks location.
  • It is interesting that Franklin County has no starbucks store




Conclusions

In our analysis, we observed that there does appear to be a strong correlation between communities with high average household incomes and Starbucks locations. In addition, starbucks locations tend to be in highly populated cities and states. This makes a lot of sense as there is a large customer base.

Questions remaining to be answered:

References

  1. Wikipedia - Starbucks Corporation
  2. Starbucks Website
  3. Starbucks Hopes 15 New Stores Will Make It Part Of The ‘DNA’ Of Low-Income Communities Of Color - Forbes.
  4. Starbucks-led coalition seeks to engage 100,000 at-risk youth
  5. Why Starbucks’ Race Together Campaign Failed
  6. Starbucks Race ‘Conversation Starters’ Are All Equally Terrible