Favorite and retweet counts from @year_progress

At the end of every year, the percentage updates from @year_progress become increasingly difficult to ignore. I had wondered for a while whether engagement grew smoothly through the year or concentrated around memorable percentages.

This post preserves the original 2020 experiment. The social network API is no longer queried during rendering: the observations were recovered from the JSON embedded in the historical Highcharter widget and are now stored locally.

library(dplyr)
library(highcharter)
library(purrr)
library(readr)
library(tidyr)

engagement <- read_csv(
  "data/year-progress-2020.csv",
  show_col_types = FALSE
) |>
  mutate(
    created_at = as.Date(created_at),
    progress_number = readr::parse_number(progress)
  ) |>
  arrange(created_at, type)

engagement |>
  slice_head(n = 8)

# A tibble: 8 × 5
  created_at progress type      count progress_number
  <date>     <chr>    <chr>     <dbl>           <dbl>
1 2020-01-04 1%       Favorites 29516               1
2 2020-01-04 1%       Retweets   5301               1
3 2020-01-08 2%       Favorites 16492               2
4 2020-01-08 2%       Retweets   2220               2
5 2020-01-12 3%       Favorites 12854               3
6 2020-01-12 3%       Retweets   1741               3
7 2020-01-15 4%       Favorites 10678               4
8 2020-01-15 4%       Retweets   1578               4

The rhythm of the year

hchart(
  engagement,
  "line",
  hcaes(created_at, count, group = type)
) |>
  hc_title(text = "Engagement with @year_progress during 2020") |>
  hc_yAxis(title = list(text = "Count")) |>
  hc_xAxis(title = list(text = NULL)) |>
  hc_tooltip(
    shared = TRUE,
    useHTML = TRUE,
    backgroundColor = "rgb(244, 246, 248)",
    borderWidth = 0,
    headerFormat = "<b>{point.key}</b><br>Progress: {point.progress}<br>"
  ) |>
  hc_credits(enabled = FALSE)

The peaks are not random. Round milestones such as 10%, 25%, 50% and 75% are easy to recognize and share. A few less conventional numbers also stand out, which is precisely what makes the pattern more interesting than a simple rise toward New Year’s Eve.

Highlighting the favorites

The original post annotated the five updates with the largest favorite count. The labels preserved the account’s tiny text visualization: 15 filled or empty blocks followed by the percentage. The final 0% belongs to January 1, 2021, when the progress bar resets for the new year.

favorite_points <- engagement |>
  filter(type == "Favorites") |>
  slice_max(count, n = 5, with_ties = FALSE) |>
  mutate(
    filled_blocks = round(15 * progress_number / 100),
    annotation_text = paste0(
      strrep("▓", filled_blocks),
      strrep("░", 15 - filled_blocks),
      " ",
      progress
    )
  )

top_annotations <- favorite_points |>
  transmute(
    x = highcharter::datetime_to_timestamp(created_at),
    y = count,
    text = annotation_text,
    point = purrr::map2(
      x,
      y,
      ~ list(x = .x, y = .y, xAxis = 0, yAxis = 0)
    )
  ) |>
  select(text, point) |>
  highcharter::list_parse()

hchart(
  engagement,
  "line",
  hcaes(created_at, count, group = type)
) |>
  hc_title(text = "Favorite and retweet counts from @year_progress in 2020") |>
  hc_tooltip(
    shared = TRUE,
    useHTML = TRUE,
    backgroundColor = "rgb(244, 246, 248)",
    borderWidth = 0,
    headerFormat = "<b>{point.key}</b><br>Progress: {point.progress}<br>"
  ) |>
  hc_annotations(list(
    labelOptions = list(
      shape = "connector",
      align = "right",
      justify = FALSE,
      crop = TRUE,
      style = list(fontSize = "0.8em", textOutline = "1px white")
    ),
    labels = top_annotations
  )) |>
  hc_yAxis(title = list(text = "Count")) |>
  hc_xAxis(
    title = list(text = NULL),
    type = "datetime",
    dateTimeLabelFormats = list(month = "%B")
  ) |>
  hc_credits(enabled = FALSE)

The result was mostly expected: people respond to landmarks. But the unusually popular intermediate percentages are a useful reminder that attention follows culture and jokes as much as it follows a calendar.

Recovery note

The CSV is a fixed historical snapshot recovered from the published chart. No account credentials, API calls or live social-media data are required to render this article.