Automatically take a waterfall graph in Mixpanel and create a user retention curve

user_retention_curve.R

# Plots a graph of retention after exporting data from Mixpanel's retention
# feature. See Thoughtbot blog: https://robots.thoughtbot.com/create-a-retention-curve-with-mixpanel-and-google-sheets
# and my blog: http://www.brycemcdonnell.com/plotting-user-retention-curves-in-R/

library(lubridate)
library(tidyr)
library(dplyr)
library(ggplot2)

##
# functions
##

# TODO - acquire through Mixpanel's API
juggleData <- function(path_to_csv) {
  # read in data
  retention <- read.csv(path_to_csv, header=TRUE, stringsAsFactors=TRUE)

  # type data correctly
  retention$start.date <- ymd(retention$start.date)

  # Transform data for graphing
  retentionv <- retention[, -2] %>% gather(t, retained, -start.date)
  retentionc <- retention[, 1:2]

  retentiong <- inner_join(retentionv, retentionc, by='start.date') %>%
    mutate(retention_percent =  retained / cohort.size,
           weeks_ago = as.integer(round(difftime(now(), start.date, units='weeks')))
           )
  retentiong
}

# plotting function plots n number of weeks depending on data set passed in
plotRetention <- function(dataSet) {
  ggplot(data=dataSet) +
    geom_line(aes(group=start.date, x=t, y=(retention_percent * 100),
                  color=weeks_ago),
              size=1
              ) +
    labs(title="Retention Over Time", y="% Retained", x="Weeks Since Event") +
    theme_bw()
}

retention <- juggleData("retention_week_2015-04-06_to_2015-07-05.csv")

# Plot all the things
plotRetention(retention)

# Plot interesting time scale
plotRetention( retention %>% filter(t != 'X0.weeks.later' & weeks_ago != 27))