Conversions

A Conversion is the definition of a valued action — "Purchase", "Signup", "Activated" — that you want to measure and optimize toward. It's the thing you define; what happens to a person when they complete it is their commercial state. One Conversion definition powers your experiments, journeys, funnel metrics, and scoring.

What Is a Conversion?

A Conversion is a specific user action that represents value. When a visitor performs that action, Apex records it as a conversion event and attributes it to the right experiment variant, journey arm, or marketing source.

Conversions are reusable. Define one once, then reference it from as many experiments and journeys as you need. Exactly one Conversion per workspace is flagged primary — experiments without an explicit objective default to it, and the headline funnel + scoring use it.

Conversion Types

A Conversion is observed from an owned event — a signal the snippet or SDK sees directly. Apex supports four owned-event types:

A Conversion can also carry platform links (Meta, Google, LinkedIn) for ad-ROAS reporting, but those attributed counts are never blended with the owned-event count and are never usable as an experiment or journey objective. See the measurement invariant below.

Creating a Conversion

The Conversion Configurator

When you create a Conversion, the configurator helps you target the right element. For click and form_submit types, you can:

• Enter a CSS selector directly if you know it
• Use the visual selector to point-and-click on the element you want to track
• Test the selector to verify it matches the right element on the page

Connecting Conversions to Experiments

An experiment optimizes toward a Metric — and a Metric references a Conversion's owned-event source. When you create or edit an experiment, you pick which Conversion defines success (its primary Metric); you can add secondary and guardrail Metrics too.

The flow works like this:

1. A visitor lands on the experiment's target page
2. Apex assigns them to control or variant_b
3. If the visitor completes the Conversion, Apex records it
4. The completion is attributed to whichever variant the visitor was assigned to

This attribution is automatic. The snippet tracks both experiment assignment and conversion completion, then connects them using the visitor's anonymous ID. Because the owned event is the only signal observable for both the exposed and holdout cohorts, only the owned-event source of a Conversion may serve as a Metric.

Tracking Custom Events

For custom_event conversions, send events from your code using the tracking snippet's API or the SDK:

// Client-side (snippet API)
window.apex?.track("video_complete", {
  videoId: "intro-demo",
  duration: 120,
});

// Client / Node (SDK)
import { init, track } from "@apex-inc/sdk";

init({ workspaceKey: "YOUR_KEY" });

track("video_complete", {
  videoId: "intro-demo",
  duration: 120,
});

For backend signals (Stripe webhooks, CRM-driven conversions), use the Server Events API:

import { sendServerEvent } from "@apex-inc/sdk";

await sendServerEvent(
  { apiKey: process.env.APEX_API_KEY! },
  { type: "purchase_completed", email: "user@example.com", data: { value: 99, currency: "USD" } },
);

The event name in your code must exactly match the eventName configured on the Conversion.

Conversion Metrics

For each Conversion, the dashboard shows:

• Total conversions — How many times the Conversion was completed
• Conversion rate — Conversions divided by unique visitors
• Per-variant breakdown — Conversion rate for each experiment variant
• Trend — How the conversion rate is changing over time

These metrics update in real-time as the snippet collects data, and every number binds to exactly one source — never a blended total.

The measurement invariant

A Conversion may carry multiple sources (an owned event plus platform links), but no surface ever counts a blended total. Every measured number binds to exactly one (source, attribution-window, counting-method) tuple:

• Used as an experiment Metric or a journey conversion event → resolves to the owned-event source only.
• Platform-attributed counts (Meta/Google/LinkedIn) are valid for ad-ROAS and breakout reporting, and forbidden as lift or calibration denominators.

This is the discipline that keeps "the numbers don't match" tickets from ever appearing.

Targets

A Target is an aggregate KPI number you commit to by a date — MRR, pipeline, "1,000 signups by Q3." Bind a Target to a single owned-event Conversion to auto-count progress from its completions; otherwise it's a manual number you update yourself. See the Targets & Conversions API for the programmatic surface.

Best Practices

• One primary Metric per experiment. Multiple co-equal Metrics make it hard to declare a winner. Track the rest as secondary or guardrail Metrics.
• Name conversions descriptively. You'll reuse them across experiments and journeys — future-you will appreciate clear names.
• Prefer specific selectors. #signup-cta is more reliable than .btn.primary which might match multiple buttons.
• Test before activating. Use the configurator to verify your selector or event name works before running an experiment against it.