Provenance — estimate narrowing
Variant: Estimated
An estimate that shows its uncertainty instead of hiding it. The figure wears an approximation sign, the dashed rule and a footnote marker that stays muted, and under it a 90% interval sits on a fixed 28–48% axis. Two sample batches land: the n count turns over first, the band narrows on a 385ms ease-in-out — it is already on screen and changing size, so it morphs rather than enters — and the interval label confirms what the band now shows only once it has settled, so the order on screen is the order of cause. The point estimate never moves; if the centre wandered while the band closed, a reader could not tell a better estimate from a different one. And it never earns the solid rule, however many samples arrive: more data buys a narrower band, not a different kind of claim. Use it for canaries, A/B readouts, forecasts and anything still sampling.
The source
A complete HTML document. Paste it into an empty .html file and it plays — no stylesheet, script, font, or image to fetch.
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1" />
<title>Provenance — estimate narrowing</title>
<style>
:root {
/* Neutrals — fixed, never remixed. */
--ivory: #faf9f5;
--paper: #ffffff;
--oat: #e3dacc;
--slate: #141413;
/* Knobs — the remix API. A knob exists iff declared here and used below. */
--hue: 38.8;
--lift: 0;
--chroma: 1;
--round: 1;
--speed: 1;
/* Accents — derived from the knobs with the family's offsets baked in,
so one hue roll moves them together. Defaults reproduce the original hex. */
--clay: oklch(calc(0.6724 + var(--lift)) calc(0.1308 * var(--chroma)) var(--hue));
--clay-d: oklch(calc(0.5797 + var(--lift)) calc(0.127 * var(--chroma)) var(--hue));
--olive: oklch(
calc(0.6118 + var(--lift)) calc(0.0713 * var(--chroma)) calc(var(--hue) + 88.3)
);
--rust: oklch(
calc(0.5408 + var(--lift)) calc(0.1357 * var(--chroma)) calc(var(--hue) - 10.3)
);
--_muted: color-mix(in srgb, var(--slate) 52%, transparent);
--_dash: color-mix(in srgb, var(--slate) 34%, transparent);
--_band: color-mix(in srgb, var(--clay) 30%, var(--paper));
--_oat-wash: color-mix(in srgb, var(--oat) 55%, var(--paper));
/* Clock — 7s: 1% is 70ms. Two sample batches land, at 30% and 58%. */
--cycle: 7s;
--t: calc(var(--cycle) / var(--speed));
--ease-out: cubic-bezier(0.23, 1, 0.32, 1);
--ease-in-out: cubic-bezier(0.77, 0, 0.175, 1);
}
* {
box-sizing: border-box;
margin: 0;
padding: 0;
}
body {
min-height: 100vh;
display: grid;
place-items: center;
background: var(--ivory);
font-family:
system-ui,
-apple-system,
'Segoe UI',
Roboto,
sans-serif;
color: var(--slate);
}
.card {
width: min(300px, calc(100vw - 32px));
padding: 14px 16px 13px;
display: grid;
gap: 10px;
border-radius: calc(12px * var(--round));
background: var(--paper);
border: 1px solid var(--oat);
animation: card var(--t) linear infinite;
}
/* The reset hides in a dark window at the wrap. */
@keyframes card {
0% {
opacity: 0;
}
2.5% {
opacity: 1;
}
95.5% {
opacity: 1;
animation-timing-function: var(--ease-out);
}
100% {
opacity: 0;
}
}
.k {
font-size: 11px;
font-weight: 700;
letter-spacing: 0.12em;
text-transform: uppercase;
color: var(--_muted);
}
.claim {
font-size: clamp(16px, 4.6vw, 19px);
font-weight: 600;
line-height: 1.35;
letter-spacing: -0.01em;
text-wrap: balance;
}
/* An estimate never earns the solid rule, however many samples arrive —
more data buys a narrower band, not a different kind of claim. */
.fig {
position: relative;
white-space: nowrap;
font-variant-numeric: tabular-nums;
}
.fig::before {
content: '';
position: absolute;
left: 0;
right: 0;
bottom: -2px;
border-bottom: 1.5px dashed var(--_dash);
}
/* Same marker as its siblings, but it stays muted: nothing has vouched. */
.ref {
font-size: 11px;
font-weight: 700;
vertical-align: super;
line-height: 0;
margin-left: 2px;
color: var(--_muted);
}
.rows {
border-top: 1px solid var(--oat);
padding-top: 9px;
display: grid;
gap: 8px;
font-size: 12px;
color: var(--_muted);
}
.row {
display: flex;
align-items: center;
gap: 7px;
}
.n {
font-size: 11px;
font-weight: 700;
}
.t {
flex: 1;
white-space: nowrap;
}
.pill {
flex: none;
padding: 2px 8px;
border-radius: 999px;
font-size: 11px;
font-weight: 600;
background: var(--_oat-wash);
}
/* The axis spans 28–48%, so the point estimate sits dead centre and
31–45 is 70% of the track. */
.ci {
position: relative;
height: 12px;
}
.ci::before {
content: '';
position: absolute;
left: 0;
right: 0;
top: 5px;
height: 2px;
border-radius: 999px;
background: var(--oat);
}
.band {
position: absolute;
top: 2px;
left: 15%;
width: 70%;
height: 8px;
border-radius: calc(2px * var(--round));
background: var(--_band);
animation: band var(--t) linear infinite;
}
/* The band is already on screen and changing size, so it morphs on the
in-out curve. 1 → 6/14 → 2/14 of its width: 31–45, 35–41, 37–39. */
@keyframes band {
0%,
30.5% {
transform: scaleX(1);
animation-timing-function: var(--ease-in-out);
}
36%,
58.5% {
transform: scaleX(0.4286);
animation-timing-function: var(--ease-in-out);
}
64%,
100% {
transform: scaleX(0.1429);
}
}
/* The point never moves. If the centre wandered as the band closed, the
reader could not tell a better estimate from a different one. */
.pt {
position: absolute;
left: calc(50% - 1px);
top: 0;
width: 2px;
height: 12px;
border-radius: 999px;
background: var(--clay-d);
}
.read {
display: flex;
justify-content: space-between;
font-variant-numeric: tabular-nums;
}
.read b {
color: var(--slate);
font-weight: 600;
}
.sw {
display: inline-grid;
}
.sw > * {
grid-area: 1 / 1;
text-align: right;
}
/* Order is causality: n turns over first (data arrived), the band
narrows, and the interval label confirms what the band now shows. */
.n1 {
animation: s1 var(--t) linear infinite;
}
.n2 {
animation: s2 var(--t) linear infinite;
}
.n3 {
animation: s3 var(--t) linear infinite;
}
.c1 {
animation: c1 var(--t) linear infinite;
}
.c2 {
animation: c2 var(--t) linear infinite;
}
.c3 {
animation: c3 var(--t) linear infinite;
}
@keyframes s1 {
0%,
30% {
opacity: 1;
animation-timing-function: ease;
}
31.5%,
100% {
opacity: 0;
}
}
@keyframes s2 {
0%,
30% {
opacity: 0;
animation-timing-function: ease;
}
31.5%,
58% {
opacity: 1;
animation-timing-function: ease;
}
59.5%,
100% {
opacity: 0;
}
}
@keyframes s3 {
0%,
58% {
opacity: 0;
animation-timing-function: ease;
}
59.5%,
100% {
opacity: 1;
}
}
@keyframes c1 {
0%,
36% {
opacity: 1;
animation-timing-function: ease;
}
37.5%,
100% {
opacity: 0;
}
}
@keyframes c2 {
0%,
36% {
opacity: 0;
animation-timing-function: ease;
}
37.5%,
64% {
opacity: 1;
animation-timing-function: ease;
}
65.5%,
100% {
opacity: 0;
}
}
@keyframes c3 {
0%,
64% {
opacity: 0;
animation-timing-function: ease;
}
65.5%,
100% {
opacity: 1;
}
}
@media (prefers-reduced-motion: reduce) {
/* One shared 1.6x clock keeps every part in phase. */
.card,
.card *,
.card *::before,
.card *::after {
animation-duration: calc(var(--t) * 1.6);
}
/* The narrowing is the reading, so the band keeps it — as three
widths crossfaded on the label’s beats instead of a morph. */
.band {
animation-name: c1;
}
.band + .band {
animation-name: c2;
}
.band + .band + .band {
animation-name: c3;
}
}
@media not (prefers-reduced-motion: reduce) {
.band + .band {
display: none;
}
}
</style>
</head>
<body>
<div
class="card"
role="img"
aria-label="An estimated figure with a dashed underline; as two batches of samples arrive, its 90% interval narrows from 31–45% to 37–39% around a point that never moves"
>
<span class="k">Claim</span>
<p class="claim">
Cache change cut p95 latency <span class="fig">≈38%</span><sup class="ref">1</sup>
</p>
<div class="rows">
<div class="row">
<span class="n">1</span>
<span class="t">Canary sample, still running</span>
<span class="pill">Estimate</span>
</div>
<div class="ci">
<span class="band"></span>
<span class="band" style="transform: scaleX(0.4286)"></span>
<span class="band" style="transform: scaleX(0.1429)"></span>
<span class="pt"></span>
</div>
<div class="read">
<span
>90% CI
<b class="sw"
><span class="c1">31–45%</span><span class="c2">35–41%</span
><span class="c3">37–39%</span></b
></span
>
<span
>n =
<b class="sw"
><span class="n1">12</span><span class="n2">48</span><span class="n3">200</span></b
></span
>
</div>
</div>
</div>
</body>
</html>
The variant family
Same idea, one axis moved. Compare them side by side before you commit.