Ridgeline Plot | Zyeon UI

status="ready" · time to interactive across 6 releases, 1.0 at the top down to 1.5, the whole distribution walking left. 1.3 is bimodal: its median barely moves, but 30% of sessions were hit by a cold-start regression — a line, a bar or a box plot all report "nothing happened". Hover or arrow through the rows; Enter pins one row's median as a dashed baseline.

Time to interactive · by release

Real user monitoring, 90th-day window, desktop sessions only

Time to interactive (ms)1,434 observations · 6 rows median peakaxis fitted to this data

1,0002,0003,000

Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that. Click a row, or press Enter on it, to pin its median as the dashed comparison line; Escape clears it. Arrow keys walk the rows.

Time to interactive · by release: every row's distribution of time to interactive in ms. Curves are Gaussian kernel density estimates with a Silverman bandwidth on one shared axis from 581.3 ms to 3,737.7 ms; medians are R-7 quantiles of the raw observations and peaks are the modes of the estimate. Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that.
Row	Observations	Minimum	Median	Peak	Maximum	Note
1.0	210	1,769 ms	2,420.5 ms	2,408.69 ms	3,105 ms	measured
1.1	240	1,629 ms	2,277.5 ms	2,242.56 ms	3,101 ms	measured
1.2	190	1,403 ms	2,079 ms	2,142.89 ms	2,658 ms	measured
1.3	214	1,478 ms	2,032 ms	1,910.31 ms	3,287 ms	measured
1.4	260	1,125 ms	1,725 ms	1,777.41 ms	2,403 ms	measured
1.5	320	1,032 ms	1,538.5 ms	1,511.61 ms	2,079 ms	measured

status="empty" · a valid payload with not one observation in it — a different sentence from "loading" and from "the fetch failed".

Time to interactive · by release

Real user monitoring, 90th-day window, desktop sessions only

No distributions yet

Curves appear once at least one group has observations.

status="loading" · five skeleton ridges, aria-hidden, no timers, and no pulsing under reduced-motion.

Time to interactive · by release

Real user monitoring, 90th-day window, desktop sessions only

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

Time to interactive · by release

Real user monitoring, 90th-day window, desktop sessions only

Couldn't load these distributions

The data source didn't respond.

Degenerate inputs in one card · fixed 0–100 axis: a label far wider than the lane (truncated, full text in the table), a single observation, 12 identical values, a row that measured nothing, exactly 2 observations, and a row spilling off both ends of the axis (negatives included) — spilled observations still count towards the median, and the footnote says how many were not painted. By row 6 chart-1..5 has wrapped back to the first colour, so colour can no longer identify anything.

Degenerate inputs · fixed 0–100 axis

Every row here is a case that makes a kernel density estimate misbehave

Completion (%)300 observations · 6 rows median peakfixed axis

050100

Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that. Click a row, or press Enter on it, to pin its median as the dashed comparison line; Escape clears it. Arrow keys walk the rows.

18 observations fell outside the drawn axis: counted in the medians, not painted. A row whose observations are all identical has no bandwidth of its own and is drawn as a narrow spike.

Degenerate inputs · fixed 0–100 axis: every row's distribution of completion in %. Curves are Gaussian kernel density estimates with a Silverman bandwidth on one shared axis from 0% to 100%; medians are R-7 quantiles of the raw observations and peaks are the modes of the estimate. Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that.
Row	Observations	Minimum	Median	Peak	Maximum	Note
Enterprise renewals, EMEA, self-serve onboarding	160	39%	62%	61.05%	90%	measured
Single observation	1	42%	42%	42.11%	42%	every observation identical
All identical	12	63%	63%	63.16%	63%	every observation identical
Nothing measured	0	—	—	—	—	no observations
Two observations	2	8%	49.5%	8.42%	91%	measured
Spills off both ends	125	-38%	48%	47.37%	134%	18 observations outside the drawn axis

heightScale="count" · the same data as density × observation count: 1.5 has 320 observations and the tightest spread, so it draws tallest; bimodal 1.3 splits its mass into two humps, has the lowest peak density, and draws shortest. The default "row" makes every ridge equally tall, which is better for comparing shape and position — at the cost of flattening exactly this difference.

Time to interactive · by release

Real user monitoring, 90th-day window, desktop sessions only

Time to interactive (ms)1,434 observations · 6 rows median peakaxis fitted to this data

1,0002,0003,000

Height is density times observations, so a row resting on more evidence is drawn taller. Click a row, or press Enter on it, to pin its median as the dashed comparison line; Escape clears it. Arrow keys walk the rows.

Time to interactive · by release: every row's distribution of time to interactive in ms. Curves are Gaussian kernel density estimates with a Silverman bandwidth on one shared axis from 581.3 ms to 3,737.7 ms; medians are R-7 quantiles of the raw observations and peaks are the modes of the estimate. Height is density times observations, so a row resting on more evidence is drawn taller.
Row	Observations	Minimum	Median	Peak	Maximum	Note
1.0	210	1,769 ms	2,420.5 ms	2,408.69 ms	3,105 ms	measured
1.1	240	1,629 ms	2,277.5 ms	2,242.56 ms	3,101 ms	measured
1.2	190	1,403 ms	2,079 ms	2,142.89 ms	2,658 ms	measured
1.3	214	1,478 ms	2,032 ms	1,910.31 ms	3,287 ms	measured
1.4	260	1,125 ms	1,725 ms	1,777.41 ms	2,403 ms	measured
1.5	320	1,032 ms	1,538.5 ms	1,511.61 ms	2,079 ms	measured

overlap={0} · no overlap, every curve in its own lane: nothing is occluded and each row can be measured, at the cost of a much taller card for the same information.

Time to interactive · by release

Real user monitoring, 90th-day window, desktop sessions only

Time to interactive (ms)1,434 observations · 6 rows median peakaxis fitted to this data

1,0002,0003,000

Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that. Click a row, or press Enter on it, to pin its median as the dashed comparison line; Escape clears it. Arrow keys walk the rows.

Time to interactive · by release: every row's distribution of time to interactive in ms. Curves are Gaussian kernel density estimates with a Silverman bandwidth on one shared axis from 581.3 ms to 3,737.7 ms; medians are R-7 quantiles of the raw observations and peaks are the modes of the estimate. Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that.
Row	Observations	Minimum	Median	Peak	Maximum	Note
1.0	210	1,769 ms	2,420.5 ms	2,408.69 ms	3,105 ms	measured
1.1	240	1,629 ms	2,277.5 ms	2,242.56 ms	3,101 ms	measured
1.2	190	1,403 ms	2,079 ms	2,142.89 ms	2,658 ms	measured
1.3	214	1,478 ms	2,032 ms	1,910.31 ms	3,287 ms	measured
1.4	260	1,125 ms	1,725 ms	1,777.41 ms	2,403 ms	measured
1.5	320	1,032 ms	1,538.5 ms	1,511.61 ms	2,079 ms	measured

A single row · with no neighbour there is no shift between rows to report and nothing to occlude; the component draws as usual and the summary sentence falls back to its single-row wording.

Checkout latency · one segment

A ridgeline of one is just a density curve — and still has to draw

Checkout latency (ms)300 observations · 1 row median peakaxis fitted to this data

5001,000

Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that. Click a row, or press Enter on it, to pin its median as the dashed comparison line; Escape clears it. Arrow keys walk the rows.

Checkout latency · one segment: every row's distribution of checkout latency in ms. Curves are Gaussian kernel density estimates with a Silverman bandwidth on one shared axis from 214.86 ms to 1,116.14 ms; medians are R-7 quantiles of the raw observations and peaks are the modes of the estimate. Every row is scaled to its own peak, so height compares shape and position, never volume — the counts beside each label carry that.
Row	Observations	Minimum	Median	Peak	Maximum	Note
All traffic	300	322 ms	636 ms	622.81 ms	1,009 ms	measured