shadcn.io is not affiliated with official shadcn/ui
Experiment Results Accordion Changelog Block for React
An accordion of four hypotheses with the verdict as an icon and a word, and in each panel an arms table and a significance meter with a tick at 95 percent.
Installation
Overview
An experiment log for React with Base UI, written for Petridish, a fictional product-experiment tool. A start-aligned header counts from the data how many of four hypotheses won, and under a band rule the hypotheses run rail to rail as one accordion. Each trigger carries the verdict as an outline badge with an icon and a word (Winner, Loser, Neutral, Inconclusive), the hypothesis with its metric, and the relative lift in words with an arrow. The first row is open: a table of the arms with visitors, conversions, rate and the change against control, and a significance meter, a bar that fills to the measured confidence with a tick at 95%. The confidence is worked out from the rates and the sample sizes, and the verdict from the confidence, so the table, the meter and the badge agree. It's built from Accordion, Badge, Progress and Guides, with CSS transitions for the entrance and your theme's tokens for every colour. For experiment pages and release notes. The tests are example data.
FAQ
How do I change the hypotheses?
Edit HYPOTHESES at the top of the file: a title, a metric, the control and variant rates in percent, the variant's name and the visitors in each arm. The conversions, the relative lift, the confidence and the verdict are computed from those.
How is the verdict chosen?
verdictOf reads the confidence: at 95% or more the variant is a Winner when it is ahead and a Loser when it is behind. Below 95%, a test with at least 20,000 visitors per arm is Neutral (it was big enough to see a difference and found none) and a smaller one is Inconclusive. Change THRESHOLD and ENOUGH to suit your team.
Which row opens first, and can several stay open?
The first one, set by FIRST_OPEN. Opening another closes it.
Add multiple to the Accordion to let several stay open.
Are the closed panels in the page's markup?
Yes. The Accordion takes hiddenUntilFound, so every closed panel stays in the markup as hidden="until-found", and the browser's find in page opens the row that holds a match.
How does it work with a keyboard and a screen reader?
Each trigger is a button inside its own h3 under the section's h2, so the headings list reads every hypothesis. Enter or Space opens a row and the button says whether it's expanded. The meter has a name that includes the confidence and the line it must cross.
Related
- Pro
A/B Test Results Changelog
Metrics and chartsA framed table of five A/B tests with the lift drawn as a bar from a zero line, a whisker for the interval, the lift in words and the decision as an icon and a word.
- Pro
Feature Adoption Curves Changelog
Metrics and chartsA reverse split with adoption curves aligned by weeks since ship, four features as buttons with their 28-day adoption and 7-day retention, the chosen curve bold.
- Pro
Accordion Changelog
Lists and timelinesA release list for release-notes software: five releases as accordion rows, each with its version, date, one-line summary and change count, the newest open on a list of typed changes, and a foot band that links to every release.
- Pro
Before After Changelog
Spotlights and mediaA compare slider for a design-handoff tool's release: one drawn editor window, its fourteen cramped toolbar icons before and three labelled groups after, under a draggable divider that also answers the arrow keys, with a short table of what changed.
- Pro
Feature Sunset Changelog
Roadmap and customersFour retiring features as ruled Collapsible rows with a four-square phase meter; each opens to its four phases, who is affected and its replacement, and a toggle group filters by area.
- Pro
Error Rate Comparison Changelog
Incidents and uptimeFour error classes as rows, each a 14-day line split by a release marker with the average after it and the change as an arrow and a sign, plus a switch for one shared scale.