Cleaned up version of https://gist.github.com/jwinternheimer/4c36ce1efe29ecd48b89

churn_exploration.R

library(data.table)
library(tidyverse) # ggplot2, tidyr, dplyr, purrr, stringr ... 
library(scales)
library(grid)
library(RColorBrewer)
###############################################################
## MOST COMMON ACTIONS DURING FIRST TWO WEEKS
###############################################################

## IMPORT AND TIDY DATA
path <- "~/Downloads/"
files <- str_c(path, c("small_user_actions.csv", "churned_users_actions.csv"))
read_data_and_assign_status <- function(file, status) {
  read.table(file, header = TRUE, sep = ",") %>% 
    mutate(user_type = status) %>% 
    rename(user_id = names(.)[1],
           action = names(.)[2],
           count = names(.)[3])
}
# read data, get better column names, add user_type 
user_actions <- map2_dfr(files, c("healthy", "churned"), read_data)

# create a summary data 
total_by_action <- user_actions %>% 
  group_by(action, user_type) %>% 
  summarise(n_users = n_distinct(user_id), 
            total_count = sum(count)) %>%
  mutate(count_per_user = total_count / n_users) 

save(total_by_action, file="~/Google Drive/total_by_action.Rda")

## TARGET ACTIONS - PLOT TOTAL ACTION COUNTS PER USER
target_actions <- map(0:4, function(i) { #<<
  total_by_action %>% 
    arrange(desc(total_count)) %>% 
    slice(1:10 + 10 * i) %>%  #<<
    pull(action)
})


## HORIZONTAL BAR PLOT OF TOP ACTIONS (BY 10)

action_bar_plot <- map(1:5, function(i) {
  total_by_action %>% 
    filter(action %in% target_actions[[i]]) %>%  #<<
    ggplot(aes(x = action, y = count_per_user, fill = user_type)) +
      geom_col(position = "dodge") + 
      coord_flip() + 
      buffer_theme() +
      labs(x = "", y = "Average Actions Per User")
})