Percentile Stat Card
A latency card with the p95 headline, a distribution track marking each percentile, and a ruled row per percentile with a magnitude bar and value.
A latency card with the p95 headline, a distribution track marking each percentile, and a ruled row per percentile with a magnitude bar and value.
The source for every Application block is included with Basic and Pro. Pick a plan to copy the code.
Already purchased? Log in
Percentile Stat Card shows the shape of a metric, not just its average. The p95 headline reads 412ms, down 38ms, over a distribution track from 0 to 1,000ms with a tick for each percentile and p95 drawn in full weight. Below it a ruled row per percentile carries a hint and value: p50 128ms median request, p75 204ms typical worst case, p95 412ms SLO boundary, p99 890ms tail from cold starts.
Percentiles are one array with a track position and a relative magnitude driving each row bar, all placed with theme tokens and no chart library. A footer grounds the sample: 2.4M requests across 3 regions, with p99 spikes traced to cold starts in us-east, so the tail number has a cause attached.
Reach for this block whenever an average would hide the tail, latency, response time, or any metric where the worst case is the one users feel, wired to your percentile query for the window.
A natural flow around it on an Application Pro page:
Before
After
One strong use is an API latency percentile card. Other distribution cards:
Tip: lead with p95 or p99, not the mean; the average is the number that lies most often about how a service actually feels.