How do I run an experiment and read its results?

1 min readLast verified 2026-06-10

Create an experiment with one utm_content value per variant, send live traffic through those links, then open the experiment to compare each variant's lift against the control.

Before you start

  • A workspace with the Attribution module active.
  • The pixel installed on your site, so clicks and conversions are being tracked.
  • Two or more versions of an ad or page you want to compare — each will get its own utm_content value.

Steps

  1. Open Experiments

    Use Sidebar → Attribution → Experiments.

  2. Click New experiment

    The New experiment dialog opens. Each variant maps to a utm_content value on your links.

  3. Name it and state your hypothesis

    Fill in Name (e.g. "Hero copy A/B") and, optionally, Hypothesis (optional) — a sentence like "Variant A will lift signup conversion by 10%."

  4. Define the variants

    Every variant has a key and a utm_content value. The first variant is the control — everything else is measured against it. You need at least two; use + Add another variant for more.

  5. Click Create as draft

    The experiment is saved with status Draft. There is currently no control to change the status — results accrue automatically from traffic matching each variant's utm_content as soon as the links are live in your ads, regardless of status.

What happens next

  • The experiment appears in the All experiments table with its status (Draft, Running, or Concluded), variant count, and start date.
  • Click its name to open Variant performance. Each variant card shows Clicks, Conversions, Revenue, and its conversion rate (conversions ÷ clicks for that variant's utm_content).
  • The first variant carries the Control badge; every other variant shows its lift vs control as a signed percentage — green means it's winning, red means it's losing.
  • Before traffic arrives you'll see No samples yet. Don't conclude on a handful of clicks — wait until the numbers are large enough that the lift stops swinging day to day. A concluded experiment keeps its results and shows the Concluded status.

Gotchas

Lift is computed relative to the first variant in the list. If you enter your challenger first and your control second, every percentage will be backwards. Symptom: your "control" shows a lift badge. Fix: create the experiment with the control variant in the first row.

Results only count traffic whose utm_content exactly matches a variant's value. A typo or a mismatched value on your UTM links means that variant sits at zero forever. If a variant shows no clicks while the ad is live, compare the link's utm_content character by character.

Variants left with an empty key or empty utm_content are silently dropped when you click Create as draft. If a variant is missing after creation, that's why.