A/B runtime + live experimentation primitives
Visual primitives for the Mufflermen workshop A/B runtime — built against real workshop experiments (Quote PDF redesign, Suburb landing CTA, Mobile dock vs sidebar). Primary metric is quote-accept rate; secondary is revenue per visitor. AU + NZ split. Visual reference only — no real backend wired.
A/B runtime + live experimentation primitives is a reusable Oak Flats Muffler Men UI primitive with documented states, accessibility expectations, theme behavior, and implementation evidence.
A/B runtime + live experimentation primitives: Visual primitives for the Mufflermen workshop A/B runtime — built against real workshop experiments (Quote PDF redesign, Suburb landing CTA, Mobile dock vs sidebar). Primary metric is quote-accept rate; secondary is revenue per visitor. AU + NZ split. Visual reference only — no real backend wired.
Experiment dashboard card
Running experiment card — status, arms (control vs treatment), winner indicator, exposed count, lift chip.
Variant editor pair
Side-by-side control vs treatment editor with live char diff. Headline + body inputs per arm.
Stat-sig calculator
Sample size + power + effect size + p-value calculator using a two-proportion z-test.
Allocation slider
Variant traffic allocation sliders with live percent total + stacked allocation bar.
Holdout audience card
Holdout group card with percent, estimated size, and exclude rules (field, operator, values).
Lift chart
Daily lift line chart with 95% confidence band, zero baseline, and tonal mid-point ticks.
Primary metric tile
Primary metric tile — value, lift delta, p-value, and significance stars (★ / ★★ / ★★★).
Segment breakdown row
Per-segment lift row — Mobile / Desktop / iOS / Android / AU / NZ with diverging bar.
Funnel impact row
Funnel step impact row with control vs treatment conversion rates and relative delta.
Early stopping card
Early-stopping rules (futility / superiority / guardrail / max duration) with armed + triggered states.
Experiment history row
Past experiment row — dates, name, final lift, outcome (shipped / iterated / killed / inconclusive), learning.
SRM warning banner
Sample ratio mismatch banner with severity, p-value, and per-arm observed vs expected splits.
Decision recommendation
Recommendation card — ship / iterate / kill / keep-running with rationale, confidence bar, expected impact.
Feature flag link row
Linked feature flag row showing environment, status (on / off / ramping / killed), and rollout %.
Full experiments
Composed cockpit — three real Mufflermen experiments (Quote PDF redesign, Suburb landing CTA, Mobile dock vs sidebar) end-to-end.