Beeswarm | Zyeon UI

status="ready" · 317 cold starts across 5 regions, and all 317 are on screen as their own dot — the count in the gutter is the number of circles in that row. The three eu-west-1 invocations past two seconds are three named instances (i-7f21ab, i-91c034, i-4d88e2), not a fat tail: hover one, or tab into the row and press End, and the readout names it. af-south-1 has 18 readings, which is exactly where a density curve starts inventing a shape it cannot support.

Cold start by region

5 groups · 317 cold starts · every one drawn as its own dot · highest i-4d88e2 at 2,610 ms in eu-west-1

one cold start
median
dots are displaced across the axis only, never along it

Cold start by region— one row per group. Every observation is drawn as its own dot, displaced only across the axis, so a dot's position along the axis is its exact value. Medians are linearly interpolated (R type 7). The lowest and highest readings are named because a beeswarm exists to make an individual outlier addressable; the readings between them are reachable by tabbing into a row and using the arrow keys.
Group	Readings	Dots drawn	Median	Lowest	Highest
us-east-1	96	96	309 ms	i-544fd4 at 191 ms	i-382a19 at 464 ms
eu-west-1	91	91	332 ms	i-0c3ec3 at 195 ms	i-4d88e2 at 2,610 ms
ap-southeast-2	72	72	514 ms	i-97ed1b at 268 ms	i-2ff7ad at 980 ms
sa-east-1	40	40	488 ms	i-1ec5d3 at 264 ms	i-1cfcef at 802 ms
af-south-1	18	18	600 ms	i-07eccd at 340 ms	i-d3b3cc at 898 ms

onPointSelect → nothing pinned

status="empty" · a valid contract with not one reading, told apart from loading and from a failed fetch.

Cold start by region

One dot per observation, displaced only across the axis

No observations to plot

A beeswarm draws one dot per row of the sample, so it needs the raw readings — not a count, an average or a percentile per category.

status="loading" · three skeleton swarms, aria-hidden, honouring prefers-reduced-motion, with no timer involved.

Cold start by region

One dot per observation, displaced only across the axis

status="error" · the fetch failed. Try again only appears when onRetry is passed, and clicking it flips the card to ready.

Cold start by region

One dot per observation, displaced only across the axis

Couldn't load these observations

The data source didn't respond.

Awkward samples · n=1 (canary) is still a dot at its exact value, not a curve. retired shard was measured and returned nothing, so it keeps its row and says so rather than quietly disappearing. cached fixture is forty runs that all landed on 250 ms — a column no dot size can fit, so that row grows instead and the other rows keep the larger dot: dropping the last twenty to make a fixed height work is the one thing this chart must never do, and counting the column gets you forty. The last label is far too long for the gutter and clamps to two lines, with the full name in the readout and the table.

Awkward samples

4 groups · 75 runs · every one drawn as its own dot · highest i-8e227a at 297 ms in provisioned-concurrency-warm-pool (eu-west-1)

one run
median
dots are displaced across the axis only, never along it

cached fixture is taller than the other rows because the swarm needed the room — dots are never dropped, and never overlapped, to make a row fit.

Awkward samples— one row per group. Every observation is drawn as its own dot, displaced only across the axis, so a dot's position along the axis is its exact value. Medians are linearly interpolated (R type 7). The lowest and highest readings are named because a beeswarm exists to make an individual outlier addressable; the readings between them are reachable by tabbing into a row and using the arrow keys.
Group	Readings	Dots drawn	Median	Lowest	Highest
canary	1	1	296 ms	i-0c41aa at 296 ms	i-0c41aa at 296 ms
retired shard	0	0	no runs recorded
cached fixture	40	40	250 ms	i-fix000 at 250 ms	i-fix039 at 250 ms
provisioned-concurrency-warm-pool (eu-west-1)	34	34	200 ms	i-e9b992 at 159 ms	i-8e227a at 297 ms

Too many to draw, said out loud · 900 + 260 invocations against the default maxPoints=400. burst traffic draws 400 of 900 as a rank sample — every k-th value of the sorted sample with both extremes kept, so the fastest and slowest invocations survive — and the card summary, the gutter (400/900), the row's accessible label and the data table all say so. steady traffic is under the cap and draws all 260. The median and the extremes are computed from every reading, never from the sample.

Invocation duration, one week

2 groups · 1,160 invocations · 660 drawn as dots (rank sample) · highest i-abbcf6 at 942 ms in burst traffic

one invocation
median
dots are displaced across the axis only, never along it

Too many readings to give every one its own dot, so these rows draw a rank sample — every k-th value of the sorted sample, both extremes kept: burst traffic 400 of 900. Every figure on this card still uses the whole sample.

Invocation duration, one week— one row per group. Every observation is drawn as its own dot, displaced only across the axis, so a dot's position along the axis is its exact value. Medians are linearly interpolated (R type 7). The lowest and highest readings are named because a beeswarm exists to make an individual outlier addressable; the readings between them are reachable by tabbing into a row and using the arrow keys.
Group	Readings	Dots drawn	Median	Lowest	Highest
burst traffic	900	400 (rank sample)	367 ms	i-2adbcd at 135 ms	i-abbcf6 at 942 ms
steady traffic	260	260	303.5 ms	i-9f858d at 175 ms	i-bf8d84 at 612 ms

maxPoints=900 on the same data · nothing is sampled now and all 1,160 dots are drawn; the sampling notice disappears because there is nothing left to disclose. rowHeight=132 buys the taller swarm the room it needs. This is the honest version of "a lot of points": denser to read and slower to lay out, which is the trade the cap exists to let you make deliberately rather than by accident.

Invocation duration, one week

2 groups · 1,160 invocations · every one drawn as its own dot · highest i-abbcf6 at 942 ms in burst traffic

one invocation
median
dots are displaced across the axis only, never along it

Invocation duration, one week— one row per group. Every observation is drawn as its own dot, displaced only across the axis, so a dot's position along the axis is its exact value. Medians are linearly interpolated (R type 7). The lowest and highest readings are named because a beeswarm exists to make an individual outlier addressable; the readings between them are reachable by tabbing into a row and using the arrow keys.
Group	Readings	Dots drawn	Median	Lowest	Highest
burst traffic	900	900	367 ms	i-2adbcd at 135 ms	i-abbcf6 at 942 ms
steady traffic	260	260	303.5 ms	i-9f858d at 175 ms	i-bf8d84 at 612 ms

Negative values + dotRadius=2.5 · revenue change per account against control, spanning −25,299 to +35,556 USD, so the axis grows a dashed rule at zero and the reader can see how many accounts fall on each side of it — a question a density silhouette answers only by eye. Smaller dots fit 168 accounts into three rows without the swarm needing to grow.

Revenue change per account vs. control

3 groups · 168 accounts · every one drawn as its own dot · highest Stracke, Cummings and Larson at 35,556 USD in Variant B

one account
median
dots are displaced across the axis only, never along it

Revenue change per account vs. control— one row per group. Every observation is drawn as its own dot, displaced only across the axis, so a dot's position along the axis is its exact value. Medians are linearly interpolated (R type 7). The lowest and highest readings are named because a beeswarm exists to make an individual outlier addressable; the readings between them are reachable by tabbing into a row and using the arrow keys.
Group	Readings	Dots drawn	Median	Lowest	Highest
Variant A	64	64	1,745 USD	Ullrich-Williamson, Wunsch and Grimes at -25,299 USD	Hermann, Mayer and Abernathy at 24,313 USD
Variant B	64	64	5,028 USD	Kohler and Sons at -19,689 USD	Stracke, Cummings and Larson at 35,556 USD
Holdout	40	40	2,229.5 USD	Padberg - Mohr-Daniel at -13,034 USD	Maggio and Sons at 16,641 USD