Bon Datatite - A Data-Driven Exploration for Recipes and Popularity

A data-driven exploration of how time, complexity, ingredients, and nutrition shape recipe popularity on Food.com.

What Makes a Recipe Popular? A Data‑Driven Look at How We Cook at Home

Why do some recipes take off while others barely get noticed? It’s a question every home cook, food blogger, and recipe platform has wondered about and one that feels surprisingly personal. We all have dishes we return to again and again, and others we scroll past without a second thought. But what actually drives that difference at scale. To explore this, I turned to a massive data set of more than 500,000 recipes published on Food.com. With recipes entered between the years 1999 and 2016 I will be manipulating variables on ingredients, cooking times, ratings, and nutritional information to examine the patterns behind recipe reviews and popularity.

Main Questions

My goal was to understand what makes a recipe resonate with home cooks and what leads to increased reviews/scores. I approached this through three guiding questions:

• Time: Do quicker recipes attract more engagement?
• Complexity: Are simpler dishes more popular, or do cooks reward effort?
• Ingredients: What are dish categories comprising of that make them popular?

The result is the infographic below — a three‑course exploration of how we cook, what we choose, and why certain recipes rise are popular.

Data

To answer these questions, I used data from Kaggle that sourced online recipe data from Food.com. This data provided recipe data from the years 1999 to 2019 and included various variables included but limited to:

• Ingredients
• Recipe Category
• Calories
• Number of Reviews

Final Infographic

Designing the Infographic

Creating this infographic was as much a design challenge as a data challenge. Below, I walk through the ten design elements that shaped my process and how each contributed to the final product.

1. Hierarchy & Visual Flow

I structured the infographic like a light dinner: a main dish and a dessert. This metaphor created a natural visual flow and helped me organize the story:

• The 30‑minute plot opens the narrative with a clear, intuitive insight.
• The ingredient sunburst act as a dish that deepen the this data analysis.
• The complexity “layered cake” plot anchors the center.

Like a dinner menu, this structure guides the viewer through the thought process to the conclusion.

2. Color

I used a warm, food‑inspired palette — terracotta, caramel, cream, and sage — inspired by Bon Appétit and Sûr La Table. These tones:

• evoke a culinary aesthetic
• differentiate plot elements without overwhelming the sense
• maintain a general cohesion across all visualizations

The palette also reinforces the editorial tone I wanted.

3. Typography

I used Copperplate Gothic Light and bold for body text and annotations to give the aesthetic of elegance and readability.

4. Annotation

Annotations were used for turning raw data into a narrative that many cooks think about - time, complexity and ingredient usage:

• “44.6% of recipes are less than 30 minutes” label highlights the dramatic drop‑off in engagement.

The annotation helps my audience understand the story without needing to decode the data.

5. Layout

I used spacing, centered titles, and consistent margins to maintain an airy, magazine‑style feel. The sunburst plot was edited in Affinity to ensure readibility and balance among the rectangular bar chart.

6. Simplicity

Each visualization focuses on a single idea:

• Time relates to engagement
• Complexity relates to rating tiers
• Ingredients relates to category identity

In regards to the sunburst plot, I intentionally avoided overplotting and limited categories (top three ingredients per category).

7. Consistency

All plots share:

• the general background tone (#F5EFE6 / #F7F2EB)
• the same font style/family
• a cohesive palette (tans/creams/caramel colors)

8. Accessibility

I used:

• high‑contrast text
• clear labels
• intuitive axis scales
• non‑color cues (stacking order, spacing)

While the palette is not strictly colorblind‑optimized, contrast remains strong and readability is preserved.

9. Storytelling

Each visualization answers a sub‑question, but together they reveal a broader pattern:

• Home cooks overwhelmingly prefer quick recipes.
• Moderate complexity is the sweet spot across rating tiers.
• Ingredient families define the identity of recipe categories with ingredients such as Garlic Cloves, Eggs, and Onion as the top ingredients.

10. Applying a DEI Lens

The data set does not directly represent people or communities in a way that raises equity concerns, but I was mindful of how I categorized cuisines. Perhaps looking into ethnic dishes can be a future analysis.

Closing Thoughts

This project taught me how messy real‑world food data can be and how much care is required to clean, parse, and interpret it responsibly. From ISO 8601 time strings to inconsistent ingredient formatting, the preprocessing shaped what was possible in the final visualizations. This data set originally comprised of over 500k observations. Through preprocessing, I was able to filter the data for this analysis which left a little under 200k observations, giving me a better look at how rich data can be through the right strainers, appliances and funnels we have in our “kitchen” of data tools.

Home cooks gravitate toward quick, moderately complex dishes built from familiar ingredients. The patterns are surprisingly consistent across categories and rated recipes. In the end, this infographic is both a data story and a culinary one. It gives us a look at the preferences and habits that shape how we cook and eat at home.

Full Foldable Code

If you would like to reference the code used in creating this infographic, you can access it below. Please note that customized color palettes and other organizational edits were made in Affinity

# Load Libaries
library(tidyverse)
library(janitor)
library(lubridate)
library(showtext)
library(plotly)

font_add_google("Playfair Display", "playfair")
font_add_google("Lato", "lato")
showtext_auto()

# Import and Clean Data

recipes_raw <- read_csv("data/recipes_reviews/recipes.csv") |>
  clean_names()

recipes_clean <- recipes_raw |>
  filter(
    !is.na(aggregated_rating),
    !is.na(review_count),
    review_count > 0,
    calories < 9000,
    !is.na(recipe_category)
  ) |>
  select(
    name, author_name, prep_time, cook_time, total_time, date_published,
    recipe_category, keywords, recipe_ingredient_quantities,
    recipe_ingredient_parts, aggregated_rating, review_count, calories,
    fat_content, saturated_fat_content, cholesterol_content,
    sodium_content, carbohydrate_content, fiber_content,
    sugar_content, protein_content, recipe_servings
  ) |>
  mutate(
    category_group = case_when(
      str_detect(tolower(recipe_category), "dessert|cookie|cake|pie|candy|ice cream|cheesecake|brownie|pudding|tart") ~ "Dessert",
      str_detect(tolower(recipe_category), "beverage|drink|smoothie|shake|punch|cocktail|margarita") ~ "Beverage",
      str_detect(tolower(recipe_category), "breakfast|brunch|oatmeal|pancake|waffle|eggs") ~ "Breakfast",
      str_detect(tolower(recipe_category), "chicken|beef|pork|lamb|seafood|fish|salmon|tuna|steak|turkey|meat|main dish|casserole|pot roast|meatloaf|meatball") ~ "Main Dish",
      str_detect(tolower(recipe_category), "side|vegetable|potato|rice|beans|salad|grain") ~ "Side Dish",
      str_detect(tolower(recipe_category), "snack|appetizer|dip|spread") ~ "Appetizer",
      str_detect(tolower(recipe_category), "soup|stew|chowder|gumbo") ~ "Soup",
      str_detect(tolower(recipe_category), "bread|biscuit|scone|muffin") ~ "Bread",
      str_detect(tolower(recipe_category), "mexican|asian|italian|greek|african|japanese|thai|spanish|german|indian|cuban|korean|vietnamese") ~ "Cuisine",
      TRUE ~ "Other"
    ),
    year_published = year(date_published),
    ingredient_count = str_count(recipe_ingredient_parts, ",") + 1
  ) |>
  filter(category_group != "Other")


# Parse ISO 8601 Time


parse_iso_time <- function(time_str) {
  if (is.na(time_str) || time_str == "" || time_str == "None") return(NA_real_)
  clean_str <- str_remove(time_str, "^PT")
  hours <- str_extract(clean_str, "\\d+(?=H)") |> as.numeric() |> replace_na(0)
  minutes <- str_extract(clean_str, "\\d+(?=M)") |> as.numeric() |> replace_na(0)
  seconds <- str_extract(clean_str, "\\d+(?=S)") |> as.numeric() |> replace_na(0)
  total <- hours * 60 + minutes + seconds / 60
  if (total == 0) return(NA_real_)
  return(total)
}

recipes_clean <- recipes_clean |>
  mutate(
    prep_minutes = map_dbl(prep_time, parse_iso_time),
    cook_minutes = map_dbl(cook_time, parse_iso_time),
    total_minutes = map_dbl(total_time, parse_iso_time),
    time_category = case_when(
      total_minutes <= 30 ~ "Quick (≤30 min)",
      total_minutes <= 60 ~ "Medium (31-60 min)",
      total_minutes <= 120 ~ "Long (1-2 hrs)",
      total_minutes > 120 ~ "Very Long (>2 hrs)",
      TRUE ~ NA_character_
    ),
    complexity = case_when(
      ingredient_count <= 5 ~ "Simple (≤5 ingredients)",
      ingredient_count <= 10 ~ "Moderate (6-10 ingredients)",
      ingredient_count > 10 ~ "Complex (>10 ingredients)",
      TRUE ~ NA_character_
    )
  )


# Fix Sugar Column

recipes_clean <- recipes_clean |>
  mutate(
    sugar_content = as.numeric(sugar_content)
  ) |>
  filter(
    sugar_content >= 0,
    sugar_content <= 500
  )


# Visualizations

# 1. 30-Minute Sweet Spot
#| eval: true
#| echo: false
#| warning: false
#| message: false
#| fig-width: 12
#| fig-height: 8


font_add_google("Playfair Display", "playfair")
font_add_google("Lato", "lato")
showtext_auto()

time_volume <- recipes_clean |>
  filter(!is.na(total_minutes), total_minutes < 180) |>
  mutate(time_bin = cut(total_minutes, breaks = seq(0, 180, 5))) |>
  count(time_bin) |>
  mutate(midpoint = as.numeric(str_extract(time_bin, "\\d+")) + 2.5)

under30 <- sum(time_volume$n[time_volume$midpoint <= 30])
total_recipes <- sum(time_volume$n)
pct_under30 <- round((under30 / total_recipes) * 100, 1)

ggplot(time_volume, aes(x = midpoint, y = n)) +
  geom_area(
    data = time_volume |> filter(midpoint <= 30),
    fill = "#C7956D", alpha = 0.85
  ) +
  geom_area(
    data = time_volume |> filter(midpoint > 30),
    fill = "#8B7355", alpha = 0.5
  ) +
  geom_vline(
    xintercept = 30,
    color = "#4A4A4A",
    linewidth = 1,
    linetype = "longdash",
    alpha = 0.4        # make the line lighter
  ) +

  # Update annotation position
  annotate(
    "text",
    x = 30, y = max(time_volume$n) * 0.97,
    label = "30 minutes",
    family = "lato",
    size = 7,
    fontface = "bold",
    color = "#2C2416",
    hjust = -0.1
  ) +

  # Annotations and styling
  annotate(
    "text", 
    x = 18, y = max(time_volume$n) * 0.80,
    label = "The 30-Minute\nSweet Spot",
    family = "playfair", 
    size = 10, 
    fontface = "bold.italic",
    color = "#2C2416",
    lineheight = 0.9
  ) +
  annotate(
    "text",
    x = 18, y = max(time_volume$n) * 0.65,
    label = "Where home cooks thrive",
    family = "lato",
    size = 6,
    color = "#5A5A5A",
    fontface = "italic"
  ) +
  annotate(
    "text",
    x = 18, y = max(time_volume$n) * 0.50,
    label = paste0(pct_under30, "% of all recipes\nare ≤ 30 minutes"),
    family = "lato",
    size = 7,
    color = "#2C2416",
    fontface = "bold"
  ) +
  annotate(
    "curve",
    x = 35, xend = 90,
    y = max(time_volume$n) * 0.55, yend = max(time_volume$n) * 0.35,
    arrow = arrow(length = unit(0.2, "cm"), type = "closed"),
    color = "#5A5A5A", 
    linewidth = 0.8,
    curvature = -0.2
  ) +
  annotate(
    "text",
    x = 100, y = max(time_volume$n) * 0.38,
    label = "Engagement\ndeclines sharply",
    family = "lato",
    size = 6,
    color = "#5A5A5A",
    fontface = "italic"
  ) +
  annotate(
    "label",
    x = 15, y = max(time_volume$n) * 0.30,
    label = paste0(scales::comma(sum(time_volume$n[time_volume$midpoint <= 30])), 
                   " recipes"),
    family = "lato",
    size = 6,
    color = "#2C2416",
    fill = "#F5EFE6",
    label.padding = unit(0.3, "lines")
  ) +

  scale_y_continuous(
    labels = scales::comma,
    expand = expansion(mult = c(0, 0.1))
  ) +
  scale_x_continuous(
    breaks = seq(0, 180, 30),
    expand = expansion(mult = c(0, 0.02))
  ) +

  labs(
    title = "The Art of Quick Cooking",
    subtitle = "Home cooks gravitate toward recipes under 30 minutes",
    x = "Total Cook Time (minutes)",
    y = "Number of Recipes",
    caption = "Analysis of 188,000 recipes from Food.com | © 2026"
  ) +

  theme_minimal(base_family = "lato") +
  theme(
    plot.background = element_rect(fill = "#F5EFE6", color = NA),
    panel.background = element_rect(fill = "#F5EFE6", color = NA),

    # remove girdlines
    panel.grid = element_blank(),

    plot.title = element_text(
      family = "playfair",
      size = 55,
      face = "bold",
      color = "#2C2416"
    ),
    plot.subtitle = element_text(
      family = "lato",
      size = 28,
      color = "#5A5A5A"
    ),
    axis.text = element_text(color = "#5A5A5A", size = 14),
    axis.title = element_text(color = "#2C2416", size = 16, face = "bold")
  )
######################################################
# 2. Complexity by Rating Tier

font_add_google("Playfair Display", "playfair")
font_add_google("Lato", "lato")
showtext_auto()

# Data prep for rating plot
df <- recipes_clean |>
  mutate(
    rating_cat = case_when(
      aggregated_rating >= 4.5 ~ "Excellent",
      aggregated_rating >= 4.0 ~ "Very Good",
      aggregated_rating >= 3.5 ~ "Good",
      TRUE ~ "Below Average"
    ),
    complexity_cat = case_when(
      ingredient_count <= 5 ~ "Simple",
      ingredient_count <= 10 ~ "Moderate",
      TRUE ~ "Complex"
    ),
    rating_cat = factor(
      rating_cat,
      levels = c("Below Average", "Good", "Very Good", "Excellent")
    )
  ) |>
  count(rating_cat, complexity_cat) |>
  group_by(rating_cat) |>
  mutate(prop = n / sum(n)) |>
  ungroup()

cake_palette <- c(
  "Simple"   = "#6D401F",
  "Moderate" = "#EDE2CC",
  "Complex"  = "#A5583A"
)

# Base plot
p <- ggplot(df, aes(x = rating_cat, y = prop, fill = complexity_cat)) +
  geom_col(
    color = "white",
    linewidth = 1,
    width = 0.7   # add spacing between bars
  ) +
  scale_y_continuous(labels = scales::percent_format()) +
  scale_fill_manual(values = cake_palette) +
  labs(
    title = "About Half of All Recipes Are Moderately Complex",
    subtitle = "This pattern holds across every rating tier",
    x = NULL,
    y = NULL,
    fill = NULL
  ) +
  theme_minimal(base_family = "lato") +
  theme(
    plot.background = element_rect(fill = "#F7F2EB", color = NA),
    panel.grid = element_blank(),
    plot.title = element_text(
      family = "playfair",
      size = 32,
      face = "bold",
      color = "#3A2E2A"
    ),
    plot.subtitle = element_text(
      family = "lato",
      size = 20,
      color = "#5A4634"
    ),
    axis.text.x = element_blank(),   # remove x-axis labels
    axis.text.y = element_text(
      family = "lato",
      size = 16,
      color = "#4A3F35"
    ),
    legend.text = element_text(
      family = "lato",
      size = 14,
      color = "#4A3F35"
    ),
    plot.margin = margin(20, 20, 20, 20)
  )

# Add left-side labels for each bar
# Convert factor positions to numeric for annotations
rating_positions <- as.numeric(levels(df$rating_cat))

p +
  annotate(
    "text",
    x = rating_positions - 0.55,   # Shift labels left of bars
    y = 0.5,                       # Vertically centered
    label = levels(df$rating_cat),
    family = "lato",
    size = 6,
    hjust = 1,
    color = "#4A3F35"
  )
##########################################################
# 3. Radial Calories Over Time

font_add_google("Playfair Display", "playfair")
font_add_google("Lato", "lato")
showtext_auto()

cal_time <- recipes_clean |>
  filter(calories > 0, calories < 8000) |>
  group_by(year_published) |>
  summarise(median_cal = median(calories, na.rm = TRUE)) |>
  filter(!is.na(year_published))

ggplot(cal_time, aes(
  x = factor(year_published),
  y = median_cal,
  fill = median_cal
)) +
  geom_col(width = 1, color = "white") +
  coord_polar() +
  scale_fill_gradient(low = "#E9C46A", high = "#C14953") +
  labs(
    title = "Calories Over Time",
    subtitle = "Median calories per year tend to be between 300 - 350 calories per serving.",
    x = NULL, y = NULL, fill = "Median Calories"
  ) +
  theme_void(base_family = "lato") +
  theme(
    plot.background = element_rect(fill = "#F5EFE6", color = NA),
    plot.title = element_text(
      family = "playfair",
      size = 26,
      face = "bold",
      hjust = 0.5
    ),
    plot.subtitle = element_text(
      family = "lato",
      size = 12,
      hjust = 0.5
    ),
    legend.position = "bottom"
  ) +
  annotate(
    "text",
    x = "2018",
    y = max(cal_time$median_cal) * 1.25,
    label = "Median calories peak in 2018–2019,\nthough late-year values include fewer recipes.",
    family = "lato",
    size = 3.5,
    color = "#5A5A5A",
    hjust = 0.5
  )

############################################################

# 4. Ingredient Sunburst
#| echo: false
#| warning: false
#| message: false

# Extract top 3 ingredients per category 
top_ing3 <- recipes_clean |>
  filter(
    !is.na(category_group),
    category_group %in% c("Main Dish", "Dessert", "Side Dish", "Soup", "Breakfast", "Bread"),
    !is.na(recipe_ingredient_parts)
  ) |>
  mutate(
    ingredients_clean = str_remove_all(recipe_ingredient_parts, '^c\\(|\\)$'),
    ingredients_clean = str_remove_all(ingredients_clean, '\\"')
  ) |>
  separate_rows(ingredients_clean, sep = ", ") |>
  mutate(
    ingredient = str_trim(str_to_lower(ingredients_clean))
  ) |>
  filter(
    ingredient != "",
    !ingredient %in% c("salt", "pepper", "water", "oil", "butter")
  ) |>
  count(category_group, ingredient, name = "ingredient_count") |>
  left_join(category_totals, by = "category_group") |>
  mutate(
    prop = ingredient_count / total_recipes,
    prop_label = scales::percent(prop, accuracy = 1)
  ) |>
  group_by(category_group) |>
  slice_max(ingredient_count, n = 3, with_ties = FALSE) |>
  ungroup()

font_add_google("Playfair Display", "playfair")
font_add_google("Lato", "lato")
showtext_auto()

# Prepare data
sunburst_df <- top_ing5 |>
  mutate(
    category_group = factor(category_group),
    ingredient = factor(ingredient)
  )

# Compute cumulative positions for each category
cat_positions <- sunburst_df |>
  group_by(category_group) |>
  summarise(total = sum(n)) |>
  mutate(
    cat_start = lag(cumsum(total), default = 0),
    cat_end = cumsum(total),
    cat_mid = (cat_start + cat_end) / 2
  )

# Merge category positions back into ingredient data
sunburst_df <- sunburst_df |>
  left_join(cat_positions, by = "category_group") |>
  group_by(category_group) |>
  mutate(
    ing_start = cat_start + cumsum(lag(n, default = 0)),
    ing_end = ing_start + n,
    ing_mid = (ing_start + ing_end) / 2
  ) |>
  ungroup()

# Inner ring (categories)
inner <- cat_positions |>
  mutate(
    xmin = cat_start,
    xmax = cat_end,
    ymin = 0,
    ymax = 1,
    label = category_group
  )

# Outer ring (ingredients)
outer <- sunburst_df |>
  mutate(
    xmin = ing_start,
    xmax = ing_end,
    ymin = 1,
    ymax = 2,
    label = ingredient
  )

# Combine for plotting
sunburst_plot_df <- bind_rows(
  inner |> mutate(type = "category"),
  outer |> mutate(type = "ingredient")
)

# Plot
ggplot() +
  # Category ring
  geom_rect(
    data = inner,
    aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax, fill = label),
    color = "#F5EFE6",
    linewidth = 0.6
  ) +
  # Ingredient ring
  geom_rect(
    data = outer,
    aes(xmin = xmin, xmax = xmax, ymin = ymin, ymax = ymax, fill = category_group),
    color = "#F5EFE6",
    linewidth = 0.4
  ) +
  # Category labels
  geom_text(
    data = inner,
    aes(x = cat_mid, y = 0.5, label = label),
    family = "playfair",
    size = 5,
    color = "#3A2E2A",
    fontface = "bold",
    angle = 0
  ) +
  # Ingredient labels 
  geom_text(
  data = outer,
  aes(
    x = ing_mid,
    y = 1.55,
    label = paste0(label, " (", prop_label, ")"),
    angle = (ing_mid / max(outer$ing_end)) * 360 - 90
  ),
  family = "lato",
  size = 4,
  color = "#4A3F35",
  hjust = 0
) +
  coord_polar(theta = "x") +
  scale_fill_manual(values = c(
  "Main Dish" = "#D97A40",
  "Dessert"   = "#E8A6C1",
  "Side Dish" = "#A3C586",
  "Soup"      = "#E2C275",
  "Breakfast" = "#F2D8A7",
  "Bread"     = "#C9A27E"
)) +
  labs(
    title = "Top 3 Ingredients by Category",
    subtitle = "A  sunburst showing the most common ingredients across recipe categories",
    fill = "Category"
  ) +
  theme_void(base_family = "lato") +
  theme(
    plot.background = element_rect(fill = "#F5EFE6", color = NA),
    legend.position = "bottom",
    plot.title = element_text(
      family = "playfair",
      size = 30,
      face = "bold",
      color = "#3A2E2A",
      hjust = 0.5
    ),
    plot.subtitle = element_text(
      family = "lato",
      size = 18,
      color = "#5A4634",
      hjust = 0.5
    )
  )


# End of Code