This entry is part of the Chart Index, the reference library for the Chart Design Field Guide.

The strip plot is the most honest chart for showing a distribution: every observation is a dot. No bins, no smoothing, no summary. The eye reads the cloud and forms its own impression of centre, spread, and outliers. For small samples — fewer than 100 observations per group — it is almost always the right form.

The jitter is the trick that makes it work. When multiple observations share a value (or near-value), their dots overlap and the count is lost. A small random horizontal nudge — jitter — separates them, and density becomes visible as a horizontal smear rather than a single dot stuck on the value line.

What it is

A strip plot maps a categorical variable to category position along one axis and a quantitative variable to position along the other. Each observation is a single point. When points would overlap, they are jittered — given a small random offset perpendicular to the value axis — to make their count visible. The form is sometimes called a one-dimensional scatter plot or univariate scatter.

The hero chart above compares checkout conversion for five experiment variants, one dot per daily cohort, twenty-eight cohorts each. The control is drawn in grey and each variant in colour; the short dark rule is the median. Every observation is on the page, so the reader can see not only that Express pay has the highest median but that it also has the widest spread, and that Saved cart is no better than control on average while being far more variable. Randomise the data and the dots slide up and down inside their jitter slots — the horizontal position carries no information and never moves.

When to use it

Strip plots are the right choice when:

  • You have small to medium samples (5–200 per group) and want every observation visible.
  • The reader's question is "what does the actual data look like?" rather than "summarise it".
  • You are comparing many groups and box plots feel too abstract.
  • You have few duplicate values — heavy overlap requires jitter or alpha blending.
  • You want to see outliers in context with the rest of the distribution.

When not to use it

  • Large samples. With 1,000+ observations per group, the cloud becomes a solid bar of dots. Use alpha blending, a bee swarm, or switch to a density / violin.
  • Many ties at exact values. Without jitter, dots stack and the cloud is unreadable. Without sufficient jitter, ties become smudges. Pick the form carefully.
  • Precise quartile reading. Strip plots show the data, not summaries. A box plot reads quartiles faster.
  • One-dimensional with no category. A single column of jittered dots is a frequency table in disguise. Use a histogram or density.

Design principles

Choose jitter width deliberately

Too little jitter and dots stack on the value line; too much and the cloud spreads across nearby categories and confuses comparison. A jitter band of around 30–50% of the category gap is the sweet spot.

Almost no jitter: ties disappear
Thirty-six scores per cohort with the jitter band nearly closed. Where several students scored the same mark the dots stack on top of each other, and a column of eight looks like a column of three.
A 30% jitter band: every score is visible
The same scores spread across a third of each band. Ties become side-by-side dots, the dense middle of each cohort reads as a wider smear, and the three columns still stay clearly separate.

Use alpha for medium samples

For samples in the 100–500 range, fully opaque dots produce a solid clump. Reduce opacity to 0.3–0.5 so overlapping dots produce visible density without losing individual points.

Opaque dots, 150 per group: a solid clump
At full opacity the centre of each strip is a flat block of colour. The reader can see the range but not where inside it the observations pile up.
The same dots at 35% opacity: density shows through
Overlapping dots now darken. The dense band in the middle of each channel reads as a deeper tone, the sparse tail above stays light, and the eye gets a density estimate for free.

Pair with summary statistics

A strip plot alone shows the data but no summary. Overlay a median line, the quartile box, or both. Most reading tasks need both what does the data look like? and where is the centre? The two figures below show the same three variants as raw points with a median rule and as a box plot summary.

Strip with a median rule: every point, plus the centre
The rule answers the summary question without hiding anything. Variant C's median sits below the control's, but its cloud also shows the two cohorts that converted far above anything the control managed.
The same variants as boxes: summary only
Faster to read, but Variant C's two exceptional cohorts have collapsed into the top whisker. Whether that is acceptable depends on whether those two days are noise or the whole point of the experiment.

Annotate the sample size

Each strip's apparent density depends on how many observations were drawn. A small note — n = 60 — under each strip, or in the subtitle when every group has the same count, stops the eye reading wide clouds as more important than narrow ones.

Use the chart for the right N

Strip plots scale poorly. Below 30 observations, every dot is meaningful. Between 30 and 200, the form is informative. Above 500, dot overlap destroys the count even with jitter. Pick the form for your N.

Highlight the group under discussion

If one strip carries the story — the winning variant, the team being reviewed, the reader's own cohort — give it the accent colour and let the others recede to grey. Few's principle of minimising visual lookup applies to point clouds as much as to bars.

One variant in colour, the rest in grey
The prose is about Express pay, so Express pay is blue. The other four variants are still there for context, but the reader's eye lands on the right column without searching a legend.

Avoid horizontal bands of guides

Heavy horizontal gridlines compete with the dot cloud. Use a single subtle baseline at the category axis and at most one or two value reference lines (mean, target). Anything more drowns the data.

Anatomy

The composition of a strip plot
An anatomical guide

A strip / jitter plot is structurally minimal: an axis, a category position per group, and a cloud of jittered dots. Optional summary marks (mean line, median, quartile band) overlay the cloud without replacing it.

  • Box plot — summary instead of raw points. Faster reading; less honest about shape.
  • Violin plot — smoothed density. Better for medium-to-large samples.
  • Bee swarm — strip plot with collision avoidance. Every dot visible, no random jitter.
  • Raincloud plot — half-violin + box + jittered strip. The hybrid for ambitious distribution displays.
  • Dot plot — one dot per category for a value. Different question entirely (summary, not distribution).

Reading list

  • Tukey, J. (1977). Exploratory Data Analysis. The case for showing the data.
  • Wilkinson, L. (1999). Dot Plots. On univariate point displays.
  • Allen, M. et al. (2019). Raincloud Plots. The strip-violin-box hybrid.
  • Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis. Practical strip-plot recipes.