Add percent change on a coord flipped `geom_bar()` chart

Question

Some data:

sub_funnel_data <- structure(list(funnelstep = structure(1:5, .Label = c("Sessions", 
"Engaged Sessions", "Subscription Funnel - Arrives on Enter Email", 
"Subscription Funnel - Arrives on Payment Details", "Direct to Paid"
), class = "factor"), N = c(92853L, 33107L, 3469L, 3149L, 113L
), Drop = c(NA, 0.356552830818606, 0.104781466155194, 0.907754396079562, 
0.0358844077484916), Rate = c(1, 0.356552830818606, 0.0373601283749583, 
0.0339138207704651, 0.00121697737283663)), row.names = c(NA, 
5L), class = c("tbl_df", "tbl", "data.frame"))

Looks like:

sub_funnel_data
# A tibble: 5 x 4
  funnelstep                                           N    Drop    Rate
* <fct>                                            <int>   <dbl>   <dbl>
1 Sessions                                         92853 NA      1      
2 Engaged Sessions                                 33107  0.357  0.357  
3 Subscription Funnel - Arrives on Enter Email      3469  0.105  0.0374 
4 Subscription Funnel - Arrives on Payment Details  3149  0.908  0.0339 
5 Direct to Paid                                     113  0.0359 0.00122

I can create a plot:

sub_funnel_data %>% 
  ggplot(aes(reorder(funnelstep, desc(funnelstep)), N)) +
  geom_bar(stat = 'identity', fill = '#39cccc') +
  coord_flip()

Looks like:

I would like to add the drop in percentage on each step to this chart. Similar to a question I posted a few years back.

I tried to modify the solution there to my needs but failed. I like the outcome of that solution, the only real difference is that now my plot is flipped.

Here's what I tried:

sub_funnel_data <- sub_funnel_data %>% mutate(End = lag(N),
                                               xpos = 1:n() - 0.5,
                                               Diff = End - N,
                                               Percent=paste(round(Diff / End * 100, 1), "%"))

Then my attempted plot: Before coord_flip():

sub_funnel_data %>% 
  ggplot(aes(reorder(funnelstep, desc(funnelstep)), N)) +
  geom_bar(stat = 'identity', fill = '#39cccc') +
  stat_summary(aes(label = scales::comma(..y..)), fun = 'sum', 
               geom = 'text', col = 'white', vjust = 1.5) +
  geom_segment(aes(x=xpos, y = End, xend = xpos, yend = N)) +
  geom_text(aes(x = xpos, y = End - Diff / 2, label = Percent), hjust = -0.2)

I can already see this is flawed. But I want to use it with coord_flip() anyway:

sub_funnel_data %>% 
  ggplot(aes(reorder(funnelstep, desc(funnelstep)), N)) +
  geom_bar(stat = 'identity', fill = '#39cccc') +
  stat_summary(aes(label = scales::comma(..y..)), fun = 'sum', 
               geom = 'text', col = 'white', vjust = 1.5) +
  geom_segment(aes(x=xpos, y = End, xend = xpos, yend = N)) +
  geom_text(aes(x = xpos, y = End - Diff / 2, label = Percent), hjust = -0.2) +
  coord_flip()

Looks like:

How can I add percentage drop off on each step to my chart? Ideally I'd like to use the Rate field in sub_funnel_data to display the corresponding percentage drop between each step. I like the lines and %ages shown at the midpoint on each chart in the linked post, I just cannot replicate it to my current use case.

Does [this](https://stackoverflow.com/questions/6017460/position-geom-text-on-dodged-barplot) help — akrun, Jul 14 '21 at 22:14
I tried adding `position = position_dodge(width = 1)` within geom_text per that solution but the plot looks a mess. I wonder if it's because I have a factor variable, funnel step. I tried changing `geom_text(aes(x = xpos, ...` to `geom_text(aes(x = reorder(xpos, desc(funnelstep)),` but similar messy result :/ — Doug Fir, Jul 14 '21 at 22:20
is this because of your labels. Can you try to wrap those labels with `str_wrap` — akrun, Jul 14 '21 at 22:22
i.e. `sub_funnel_data %>% mutate(funnelstep = str_wrap(funnelstep, width = 10)) %>%ggplot` — akrun, Jul 14 '21 at 22:23
str_wrap did make the plot look slightly better before coord_flip(). BUt the plots still fundamentally off. Yes, the first item in Rate will be NA since it divides each row based on the lag of numeric column N. I.e. it's supposed to show the % that move from ecah step starting from Sessions, through Engaged Sessions... all the way to the end point Direct to Paid — Doug Fir, Jul 14 '21 at 22:26

jared_mamrot · Accepted Answer · 2021-07-15T01:35:33.897

If you reorder your aes() variables (in this case using rev()) and make some tweaks it looks ok, e.g.

library(tidyverse)

sub_funnel_data <- structure(list(funnelstep = structure(1:5, .Label = c("Sessions", 
                                                                         "Engaged Sessions", "Subscription Funnel - Arrives on Enter Email", 
                                                                         "Subscription Funnel - Arrives on Payment Details", "Direct to Paid"), 
                                                         class = "factor"), N = c(92853L, 33107L, 3469L, 3149L, 113L),
                                  Drop = c(NA, 0.356552830818606, 0.104781466155194, 0.907754396079562, 
                                           0.0358844077484916), Rate = c(1, 0.356552830818606, 0.0373601283749583, 
                                                                         0.0339138207704651, 0.00121697737283663)),
                             row.names = c(NA, 5L), class = c("tbl_df", "tbl", "data.frame"))

sub_funnel_data %>% mutate(End = lag(N),
                           xpos = 1:n() + 0.5,
                           Diff = End - N,
                           Percent=paste(round(Diff / End * 100, 1), "%")) %>% 
  ggplot(aes(x = reorder(funnelstep, desc(funnelstep)), y = N)) +
  geom_bar(stat = 'identity', fill = '#39cccc') +
  stat_summary(aes(label = scales::comma(..y..)),
               fun = "sum", geom = "text",
               col = "grey25", vjust = 0.5, hjust = 1) +
  geom_segment(aes(x = rev(xpos), y = End, xend = rev(xpos), yend = N)) +
  geom_text(aes(x = rev(xpos - 0.1), y = End - Diff / 1.5, label = Percent), hjust = -0.2) +
  coord_flip()

This is an alternative including the 'Rate' variable:

sub_funnel_data %>% mutate(End = lag(N),
                           xpos = 1:n() + 0.5,
                           Diff = End - N,
                           Percent = paste("\u2190 ", round(Diff / End * 100, 1), "% drop", sep = ""),
                           Rate = paste("(", round(Rate * 100, 1), "%", " of total Sessions)", sep = "")) %>% 
  ggplot(aes(x = reorder(funnelstep, desc(funnelstep)), y = N)) +
  geom_bar(stat = 'identity', fill = '#39cccc') +
  stat_summary(aes(label = paste("n = ", scales::comma(..y..), sep = "")),
               fun = "sum", geom = "text",
               col = "black", vjust = 0,
               hjust = -0.05, size = 4) +
  geom_segment(aes(x = rev(xpos), y = End, xend = rev(xpos), yend = N * 1.1)) +
  geom_text(aes(x = rev(xpos - 0.1), y = End - Diff / 1, label = Percent), hjust = -0.2) +
  geom_text(aes(x = rev(xpos - 0.7), y = End - Diff, label = Rate), color = "black", hjust = 0) +
  coord_flip()

Do either of these solve your problem?

You're welcome - I learned a lot trying to solve your problem - thanks for providing all the required details in your question :) — jared_mamrot, Jul 15 '21 at 23:17

Add percent change on a coord flipped `geom_bar()` chart

1 Answers1