Marketing Atlas
한국어
% Conversion optimization

Experiment knowledge graph

실험 지식 그래프

Trace connections among experiments, gates, code paths, and conversion metrics.

···
html
<div class="graph-demo"><div class="graph-head">TRACE THE DROP <b>CHECKOUT −6%</b></div><div class="graph-nodes"><div class="graph-node" data-node="0">METRIC<small>COMPLETE</small></div><i></i><div class="graph-node" data-node="1">EXPERIMENT<small>VARIANT B</small></div><i></i><div class="graph-node" data-node="2">GATE<small>NEW FLOW</small></div><i></i><div class="graph-node" data-node="3">CODE<small>PAYMENT</small></div></div><div class="graph-caption" id="graph-caption">METRIC → EXPERIMENT → GATE → CODE</div></div>
css
.graph-demo{width:min(92%,660px);font:700 clamp(10px,3.2vmin,15px)/1.1 sans-serif}.graph-head{display:flex;justify-content:space-between;gap:4px;margin-bottom:clamp(13px,3vmin,25px)}.graph-head b{color:var(--accent)}.graph-nodes{display:flex;align-items:center}.graph-node{border:1px solid var(--fg);background:var(--surface);padding:clamp(6px,1.6vmin,14px) clamp(3px,1vmin,10px);text-align:center;min-width:0;flex:1;transition:background .3s,color .3s}.graph-node.active{background:var(--accent);color:white;border-color:var(--accent)}.graph-node small{display:block;margin-top:5px;font-size:clamp(10px,3.2vmin,12px)}.graph-nodes i{display:block;width:clamp(5px,3vmin,24px);height:2px;background:var(--accent);flex:none}.graph-caption{color:var(--muted);text-align:center;margin-top:clamp(13px,3vmin,26px)}
js
const nodes=[...document.querySelectorAll('.graph-node')];let active=0;function draw(){nodes.forEach((node,index)=>node.classList.toggle('active',index<=active));active=(active+1)%4}draw();setInterval(draw,650);
···

AI conversion diagnosisAI 전환 장애 진단

Combine drop-off, errors, and replay to narrow likely causes.

···

Behavior-linked experiments행동 데이터 연동 실험

Read funnel, replay, and survey signals beside experiment results.

···

Natural-language experiment search자연어 실험 이력 검색

Find past tests and outcomes with a question before repeating work.

···

A/B testingA/B 테스트

Randomly assign two versions to compare their effect on conversion.

An experiment knowledge graph links results to the feature gates, code paths, events, and metrics that produced them. In 2026, Statsig described making these relationships explicit to narrow an investigation when a metric changes.

The demo traces a checkout completion drop back through an experiment variant and gate to a checkout path. Each link identifies where to investigate.

A graph edge does not prove causation. Check actual exposures, event definitions, and error logs, and look for missing connections.

When to use

Use it when a metric moves during a test and the relevant change or code path is unclear.

Open as page ↗