Interactive visualisation with Plotly in R

Interactivity is useful when it helps a reader inspect values, compare subsets or navigate dense data. It is not automatically an improvement over a clear static figure. Plotly’s R package creates browser-based graphics using Plotly.js and the htmlwidgets framework.

Select Run R above a code block to execute it with R in the browser. The resulting Plotly widget is then handed to Plotly.js and rendered as a genuinely interactive chart beneath the code. The first run is slower because the browser must load R, the packages and the JavaScript renderer. PDF and printed copies still need a static alternative.

Two routes into Plotly

1. plot_ly() builds a Plotly figure directly.
2. ggplotly() converts a ggplot object into an interactive widget.

Use plot_ly() when interaction is central and you want direct control over traces. Use ggplotly() when an existing ggplot already expresses the figure well and you need modest interactivity.

Your first direct figure

library(plotly)

plot_ly(
  data = mtcars,
  x = ~wt,
  y = ~mpg,
  type = "scatter",
  mode = "markers"
)

The tilde tells Plotly to evaluate a variable inside data. A trace is one visual layer or series.

Map variables and write useful hover text

Default hover labels often expose internal variable names or irrelevant precision. Construct the text deliberately.

cars <- transform(
  mtcars,
  model = row.names(mtcars),
  transmission = ifelse(am == 1, "Manual", "Automatic")
)

plot_ly(
  cars,
  x = ~wt,
  y = ~mpg,
  color = ~transmission,
  colors = c("#6e2538", "#214c57"),
  text = ~paste0(
    "<b>", model, "</b>",
    "<br>Weight: ", round(wt, 2), " (1,000 lb)",
    "<br>Fuel economy: ", mpg, " mpg"
  ),
  hoverinfo = "text",
  type = "scatter",
  mode = "markers",
  marker = list(size = 10, opacity = 0.8)
)

Avoid placing confidential identifiers in hover text. They remain present in the HTML widget even when not immediately visible.

Titles, axes and layout

figure <- plot_ly(cars, x = ~wt, y = ~mpg, type = "scatter", mode = "markers") |>
  layout(
    title = list(text = "Fuel economy decreases as vehicle weight rises", x = 0.05),
    xaxis = list(title = "Weight (1,000 lb)", zeroline = FALSE),
    yaxis = list(title = "Fuel economy (mpg)", zeroline = FALSE),
    font = list(family = "Ubuntu, Arial, sans-serif", color = "#1c1918"),
    paper_bgcolor = "#f7f4ee",
    plot_bgcolor = "#f7f4ee",
    margin = list(l = 70, r = 25, t = 70, b = 65)
  )

figure

Plotly layout objects are nested lists. Use browser inspection and the official figure reference when you need an unfamiliar attribute.

Add traces

automatic <- subset(cars, transmission == "Automatic")
manual <- subset(cars, transmission == "Manual")

plot_ly() |>
  add_markers(
    data = automatic, x = ~wt, y = ~mpg,
    name = "Automatic", marker = list(color = "#6e2538")
  ) |>
  add_markers(
    data = manual, x = ~wt, y = ~mpg,
    name = "Manual", marker = list(color = "#214c57")
  ) |>
  layout(
    xaxis = list(title = "Weight (1,000 lb)"),
    yaxis = list(title = "Fuel economy (mpg)"),
    legend = list(orientation = "h", x = 0, y = 1.1)
  )

Layering explicit traces is useful when groups need different markers, hover templates or visibility controls.

Convert a ggplot

library(ggplot2)
library(plotly)

static_plot <- ggplot(
  cars,
  aes(
    x = wt,
    y = mpg,
    colour = transmission,
    text = paste0(model, "<br>", mpg, " mpg")
  )
) +
  geom_point(size = 3, alpha = 0.8) +
  scale_colour_manual(values = c("#6e2538", "#214c57")) +
  labs(x = "Weight (1,000 lb)", y = "Fuel economy (mpg)", colour = NULL) +
  theme_minimal(base_family = "Ubuntu")

ggplotly(static_plot, tooltip = "text")

ggplotly() translates rather than screenshots the plot. Some ggplot extensions and theme details do not translate exactly. Inspect the widget and adjust with Plotly’s layout() where necessary.

Lines and time series

set.seed(42)
weekly <- data.frame(
  week = rep(seq.Date(as.Date("2026-01-05"), by = "week", length.out = 12), 2),
  group = rep(c("A", "B"), each = 12),
  value = c(cumsum(rnorm(12, 0.4, 1)), cumsum(rnorm(12, 0.2, 1)))
)

plot_ly(
  weekly,
  x = ~week,
  y = ~value,
  color = ~group,
  colors = c("#6e2538", "#214c57"),
  type = "scatter",
  mode = "lines+markers"
)

Sort data by time within group before drawing lines. Otherwise a trace can connect observations in an unintended order.

Bars, distributions and uncertainty

estimates <- data.frame(
  outcome = c("Functional", "Perioperative", "Patient-reported"),
  estimate = c(0.72, 0.61, 0.78),
  lower = c(0.65, 0.54, 0.71),
  upper = c(0.79, 0.68, 0.85)
)

plot_ly(
  estimates,
  x = ~estimate,
  y = ~reorder(outcome, estimate),
  type = "scatter",
  mode = "markers",
  marker = list(color = "#6e2538", size = 10),
  error_x = list(
    type = "data",
    symmetric = FALSE,
    array = ~upper - estimate,
    arrayminus = ~estimate - lower
  )
) |>
  layout(xaxis = list(title = "Estimate"), yaxis = list(title = ""))

Hover does not replace visible uncertainty. A reader should not have to discover every confidence interval one point at a time.

Facets and subplots

p1 <- plot_ly(cars, x = ~wt, y = ~mpg, type = "scatter", mode = "markers")
p2 <- plot_ly(cars, x = ~hp, y = ~mpg, type = "scatter", mode = "markers")

subplot(p1, p2, shareY = TRUE, titleX = TRUE, titleY = TRUE) |>
  layout(showlegend = FALSE)

Small multiples are often easier to compare than asking readers to toggle many traces. Preserve common scales when direct comparison matters.

Control the mode bar

figure |>
  config(
    displaylogo = FALSE,
    responsive = TRUE,
    modeBarButtonsToRemove = c("lasso2d", "select2d")
  )

Remove controls that do not serve the task, but retain zoom and image export when users genuinely need them. Avoid disabling interaction merely to make the chart look cleaner.

Save and embed a widget

htmlwidgets::saveWidget(
  figure,
  file = "fuel-economy.html",
  selfcontained = TRUE
)

A self-contained file is easy to share but can be large because its JavaScript dependencies are embedded. In Quarto, placing the widget as the final expression in a code cell usually renders it directly into HTML output.

Accessibility and fallback output

Interactive charts require additional care:

• include an informative title, axis labels and units;
• provide a prose summary of the principal pattern;
• avoid depending on colour alone;
• keep hover targets large enough to acquire;
• limit decimal precision;
• provide a table or static figure where the exact values matter;
• test keyboard behaviour and zoomed layouts; and
• do not use interactivity to conceal an overcrowded figure.

PDF and Word cannot preserve an HTML widget’s interaction. Create a static ggplot or image for non-HTML output.

Performance

Thousands of SVG points can become slow. Options include aggregation, sampling, hexagonal binning, or WebGL traces such as scattergl. Sampling must be transparent and should preserve the pattern relevant to the question.

plot_ly(
  data = cars,
  x = ~wt,
  y = ~mpg,
  type = "scattergl",
  mode = "markers"
)

Common mistakes

• Using hover labels instead of clear visible labels.
• Adding animation without a time or state question.
• Letting users compare values across changing axes.
• Embedding identifiable information in the widget source.
• Converting a complex ggplot and assuming every layer survived correctly.
• Publishing only an interactive output when the report also needs PDF or print.
• Using three-dimensional perspective for ordinary comparisons.

Official references

• Plotly for R
• Plotly R figure reference
• Displaying Plotly figures