AirBNB

This is the second installment in the Company Reviews Series.

Please refer to this post for scraping code

Data Analysis

As always, let’s load the goodies and create theme for GGPLOT2.

##### Load Goodies #####
library(tidyverse)
library(stringr)
library(tidytext)
library(ggraph)
library(igraph)

##### Theme #####
theme_moma <- function(base_size = 12, base_family = "Helvetica") {
  theme(
    plot.background = element_rect(fill = "#F7F6ED"),
    legend.key = element_rect(fill = "#F7F6ED"),
    legend.background = element_rect(fill = "#F7F6ED"),
    panel.background = element_rect(fill = "#F7F6ED"),
    panel.border = element_rect(colour = "black", fill = NA, linetype = "dashed"),
    panel.grid.minor = element_line(colour = "#7F7F7F", linetype = "dotted"),
    panel.grid.major = element_line(colour = "#7F7F7F", linetype = "dotted")
  )
}

##### Load Goodies #####

library(tidyverse)

library(stringr)

library(tidytext)

library(ggraph)

library(igraph)

##### Theme #####

theme_moma <- function(base_size = 12, base_family = "Helvetica") {

theme(

plot.background = element_rect(fill = "#F7F6ED"),

legend.key = element_rect(fill = "#F7F6ED"),

legend.background = element_rect(fill = "#F7F6ED"),

panel.background = element_rect(fill = "#F7F6ED"),

panel.border = element_rect(colour = "black", fill = NA, linetype = "dashed"),

panel.grid.minor = element_line(colour = "#7F7F7F", linetype = "dotted"),

panel.grid.major = element_line(colour = "#7F7F7F", linetype = "dotted")

)

}

Next, we will remove the first row when the container was set up before scraping.

##### Remove First Row #####
data <- data %>%
  filter(Summary != "xx")

##### Remove First Row #####

data <- data %>%

filter(Summary != "xx")

AirBNB doesn’t have the date issue as Microsoft. It seems like Glassdoor changed the way they reported date sometime prior to 2014. So, converting the date for AirBNB is very simple.

##### Date Management #####
data2 <- data
data2$Posted_Date <- as.Date(data2$Posted_Date, format = " %b %d, %Y")

##### Date Management #####

data2 <- data

data2$Posted_Date <- as.Date(data2$Posted_Date, format = " %b %d, %Y")

Then we separate the columns into two.

##### Separating Employee Type and Locations #####
data2 <- data2%>%
  separate(col = Title, into = c("Employee_Type","Location"),sep = ' in ') %>%
  separate(col = Employee_Type, into = c("Employee_Type", "Title"), sep = 'ee - ')

data2$Employee_Type <- str_replace_all(data2$Employee_Type, "Employ","Employee")

##### Separating Employee Type and Locations #####

data2 <- data2%>%

separate(col = Title, into = c("Employee_Type","Location"),sep = ' in ') %>%

separate(col = Employee_Type, into = c("Employee_Type", "Title"), sep = 'ee - ')

data2$Employee_Type <- str_replace_all(data2$Employee_Type, "Employ","Employee")

Next is text processing.

##### Munging with the Words #####
#Separating Employee Type and Locations
data2 <- data2%>%
  separate(col = Title, into = c("Employee_Type","Location"),sep = ' in ') %>%
  separate(col = Employee_Type, into = c("Employee_Type", "Title"), sep = 'ee - ')

data2$Employee_Type <- str_replace_all(data2$Employee_Type, "Employ","Employee")

##### Creating New Variables for Processing #####
data2<- data2 %>%
  mutate(Summary_2 = Summary,
         Pros_2 = Pros,
         Cons_2 = Cons)

##### Change to lower #####
data2$Summary_2 <- str_replace_all(data2$Summary_2, "[:alpha:]",tolower)
data2$Pros_2 <- str_replace_all(data2$Pros_2, "[:alpha:]",tolower)
data2$Cons_2 <- str_replace_all(data2$Cons_2, "[:alpha:]",tolower)

##### Remove Punct #####
data2$Summary_2 <- str_replace_all(data2$Summary_2, "[:punct:]","")
data2$Pros_2 <- str_replace_all(data2$Pros_2, "[:punct:]","")
data2$Cons_2 <- str_replace_all(data2$Cons_2, "[:punct:]","")

##### Remove Numbers #####
data2$Summary_2 <- str_replace_all(data2$Summary_2, "[:digit:]","")
data2$Pros_2 <- str_replace_all(data2$Pros_2, "[:digit:]","")
data2$Cons_2 <- str_replace_all(data2$Cons_2, "[:digit:]","")

##### Company Name #####
name <- array(c("airbnb"))

for (i in 1:nrow(name)){
  print(i)
  for (j in 8:ncol(data2)) {
    data2[[j]] <- str_replace_all(data2[[j]],name[[i]],"company")
    print(j)
  }
}

##### Work Life Balance & Other Words #####
worklife <- array(c("work life balance", "work-life balance",
                    "work/life balance", "work life", "work&life",
                    "worklife"))

coupon <- array(c("coupons","coupon"))
credit <- array(c("credits","credit"))

for (i in 1:nrow(worklife)){
  print(i)
  for (j in 8:ncol(data2)) {
    data2[[j]] <- str_replace_all(data2[[j]],worklife[[i]],"wlb")
  }
}

for (i in 1:nrow(coupon)){
  print(i)
  for (j in 8:ncol(data2)) {
    data2[[j]] <- str_replace_all(data2[[j]],coupon[[i]],"coupon")
    data2[[j]] <- str_replace_all(data2[[j]],credit[[i]],"credit")
  }
}

##### Munging with the Words #####

#Separating Employee Type and Locations

data2 <- data2%>%

separate(col = Title, into = c("Employee_Type","Location"),sep = ' in ') %>%

separate(col = Employee_Type, into = c("Employee_Type", "Title"), sep = 'ee - ')

data2$Employee_Type <- str_replace_all(data2$Employee_Type, "Employ","Employee")

##### Creating New Variables for Processing #####

data2<- data2 %>%

mutate(Summary_2 = Summary,

Pros_2 = Pros,

Cons_2 = Cons)

##### Change to lower #####

data2$Summary_2 <- str_replace_all(data2$Summary_2, "[:alpha:]",tolower)

data2$Pros_2 <- str_replace_all(data2$Pros_2, "[:alpha:]",tolower)

data2$Cons_2 <- str_replace_all(data2$Cons_2, "[:alpha:]",tolower)

##### Remove Punct #####

data2$Summary_2 <- str_replace_all(data2$Summary_2, "[:punct:]","")

data2$Pros_2 <- str_replace_all(data2$Pros_2, "[:punct:]","")

data2$Cons_2 <- str_replace_all(data2$Cons_2, "[:punct:]","")

##### Remove Numbers #####

data2$Summary_2 <- str_replace_all(data2$Summary_2, "[:digit:]","")

data2$Pros_2 <- str_replace_all(data2$Pros_2, "[:digit:]","")

data2$Cons_2 <- str_replace_all(data2$Cons_2, "[:digit:]","")

##### Company Name #####

name <- array(c("airbnb"))

for (i in 1:nrow(name)){

print(i)

for (j in 8:ncol(data2)) {

data2[[j]] <- str_replace_all(data2[[j]],name[[i]],"company")

print(j)

}

##### Work Life Balance & Other Words #####

worklife <- array(c("work life balance", "work-life balance",

"work/life balance", "work life", "work&life",

"worklife"))

coupon <- array(c("coupons","coupon"))

credit <- array(c("credits","credit"))

for (i in 1:nrow(worklife)){

print(i)

for (j in 8:ncol(data2)) {

data2[[j]] <- str_replace_all(data2[[j]],worklife[[i]],"wlb")

}

for (i in 1:nrow(coupon)){

print(i)

for (j in 8:ncol(data2)) {

data2[[j]] <- str_replace_all(data2[[j]],coupon[[i]],"coupon")

data2[[j]] <- str_replace_all(data2[[j]],credit[[i]],"credit")

}

As AirBNB is pretty new, there are that many reviews. So I simply scraped the entire reviews which include international reviews. Let’s categorize them into “Domestic” and “International.”

##### Domestic or International #####
data2 <- data2 %>%
  mutate(Location_2 = ifelse(str_detect(Location, "[(]")==TRUE,"International","Domestic"))

##### Domestic or International #####

data2 <- data2 %>%

mutate(Location_2 = ifelse(str_detect(Location, "[(]")==TRUE,"International","Domestic"))

Let’s see the number of reviews by month.

data2 %>%
  select(Posted_Date, Location_2) %>%
  arrange(Posted_Date) %>%
  mutate(Month = format(Posted_Date, "%m")) %>%
  mutate(Year = format(Posted_Date, "%y")) %>%
  select(Month, Year, Location_2) %>%
  ggplot(aes(x = Month, fill = Location_2)) + 
  geom_bar(stat = 'count', alpha =0.7) +
  theme_moma() +
  theme(legend.position = "bottom", axis.text.x = element_text(angle = 90, size = 7)) +
  facet_grid(.~Year)

data2 %>%

select(Posted_Date, Location_2) %>%

arrange(Posted_Date) %>%

mutate(Month = format(Posted_Date, "%m")) %>%

mutate(Year = format(Posted_Date, "%y")) %>%

select(Month, Year, Location_2) %>%

ggplot(aes(x = Month, fill = Location_2)) +

geom_bar(stat = 'count', alpha =0.7) +

theme_moma() +

theme(legend.position = "bottom", axis.text.x = element_text(angle = 90, size = 7)) +

facet_grid(.~Year)

The review just only started in 2014; each month there has been only between 10 and 20 reviews. Well, I’d think the data is on the few sides. So it won’t be necessary to separate the reviews by Group. I’ll just simply process the entire dataset for ngram chart.

Let’s start with Pros

#### Overall - Pros #####
data_pros <- data2 %>% select(Pros_2) %>% 
  unnest_tokens(words, Pros_2, token = 'ngrams',n = 2)

data_pros_split <- data_pros %>%
  separate(words, c("from","to",sep = " ")) %>%
  select(1:2)

#Step 3: Remove stopwords
data_pros_clean <- data_pros_split %>%
  filter(!from %in% stop_words$word) %>%
  filter(!to %in% stop_words$word)

#Step 4: Count
data_pros_counts <- data_pros_clean %>% 
  count(from, to)

pros_bigram <- data_pros_counts %>%
  filter(n > 5) %>%
  graph_from_data_frame()

arrow_control <- grid::arrow(type = "closed", length = unit(.15, "inches"))
Pros_chart <- ggraph(pros_bigram) +
  geom_edge_link(aes(edge_alpha = n), show.legend = FALSE,
                 arrow = arrow_control) +
  geom_node_point(color = "lightgreen", size = 3) +
  geom_node_text(aes(label = name), vjust = 1, hjust = 1) +
  theme_void() + theme( plot.background = element_rect(fill = "#F7F6ED")) +
  ggtitle("Pros")
Pros_chart

#### Overall - Pros #####

data_pros <- data2 %>% select(Pros_2) %>%

unnest_tokens(words, Pros_2, token = 'ngrams',n = 2)

data_pros_split <- data_pros %>%

separate(words, c("from","to",sep = " ")) %>%

select(1:2)

#Step 3: Remove stopwords

data_pros_clean <- data_pros_split %>%

filter(!from %in% stop_words$word) %>%

filter(!to %in% stop_words$word)

#Step 4: Count

data_pros_counts <- data_pros_clean %>%

count(from, to)

pros_bigram <- data_pros_counts %>%

filter(n > 5) %>%

graph_from_data_frame()

arrow_control <- grid::arrow(type = "closed", length = unit(.15, "inches"))

Pros_chart <- ggraph(pros_bigram) +

geom_edge_link(aes(edge_alpha = n), show.legend = FALSE,

arrow = arrow_control) +

geom_node_point(color = "lightgreen", size = 3) +

geom_node_text(aes(label = name), vjust = 1, hjust = 1) +

theme_void() + theme( plot.background = element_rect(fill = "#F7F6ED")) +

ggtitle("Pros")

Pros_chart

“Travel Coupons,” “Travel Credits,” “Smart People,” “Free Foods,” “Health Insurance.” Okay it’s about the same theme as Microsoft. Perks are important. Smart people love working with smart people. The differences are the Travel Credits… and most importantly, compensation. Hm.

Could compensation be in the Cons?

#### Overall - Cons #####
data_cons <- data2 %>% select(Cons_2) %>% 
  unnest_tokens(words, Cons_2, token = 'ngrams',n = 2)

data_cons_split <- data_cons %>%
  separate(words, c("from","to",sep = " ")) %>%
  select(1:2)

#Step 3: Remove stopwords
data_cons_clean <- data_cons_split %>%
  filter(!from %in% stop_words$word) %>%
  filter(!to %in% stop_words$word)

#Step 4: Count
data_cons_counts <- data_cons_clean %>% 
  count(from, to)

cons_bigram <- data_cons_counts %>%
  filter(n > 5) %>%
  graph_from_data_frame()

arrow_control <- grid::arrow(type = "closed", length = unit(.15, "inches"))
cons_chart <- ggraph(cons_bigram) +
  geom_edge_link(aes(edge_alpha = n), show.legend = FALSE,
                 arrow = arrow_control) +
  geom_node_point(color = "lightgreen", size = 3) +
  geom_node_text(aes(label = name), vjust = 1, hjust = 1) +
  theme_void() + theme( plot.background = element_rect(fill = "#F7F6ED")) +
  ggtitle("cons")
cons_chart

#### Overall - Cons #####

data_cons <- data2 %>% select(Cons_2) %>%

unnest_tokens(words, Cons_2, token = 'ngrams',n = 2)

data_cons_split <- data_cons %>%

separate(words, c("from","to",sep = " ")) %>%

select(1:2)

#Step 3: Remove stopwords

data_cons_clean <- data_cons_split %>%

filter(!from %in% stop_words$word) %>%

filter(!to %in% stop_words$word)

#Step 4: Count

data_cons_counts <- data_cons_clean %>%

count(from, to)

cons_bigram <- data_cons_counts %>%

filter(n > 5) %>%

graph_from_data_frame()

arrow_control <- grid::arrow(type = "closed", length = unit(.15, "inches"))

cons_chart <- ggraph(cons_bigram) +

geom_edge_link(aes(edge_alpha = n), show.legend = FALSE,

arrow = arrow_control) +

geom_node_point(color = "lightgreen", size = 3) +

geom_node_text(aes(label = name), vjust = 1, hjust = 1) +

theme_void() + theme( plot.background = element_rect(fill = "#F7F6ED")) +

ggtitle("cons")

cons_chart

Okay, “Hyper Growth,” “Career Development,” “Middle Management,” “Upper Managment.” I’m not surprised by those comments. But “Growing Pains”?

We need to dig a bit deeper in that. Let’s filter the only Cons_2 with the phrase.

##### Growing Pains? #####
growing_pain <- data2 %>%
  select(Cons_2) %>%
  filter(str_detect(Cons_2, "growing pains"))

dim(growing_pain)

##### Growing Pains? #####

growing_pain <- data2 %>%

select(Cons_2) %>%

filter(str_detect(Cons_2, "growing pains"))

dim(growing_pain)

> dim(growing_pain)
[1] 27  1

1 2	> dim(growing_pain) [1] 27 1

There are 27 comments with “Growing Pains.” I went to take a glimpse at the dataframe and found that mostly was about the hyper growth of the company. I guess that wore down to the culture. But hey at least compensation is not in the cons. So at least they are satisfied with their salary.