Run a website A/B test
Create a controlled experiment, validate both variants, and interpret results without stopping too early.
See it in Leado
What you’ll achieve
- A controlled, reproducible experiment with stable assignment and a predefined primary decision metric.
What you’ll need
- An installed Leado pixel
- A measurable conversion event
Follow the workflow
-
1
Define one measurable hypothesis
Before touching the page, write the target audience, one proposed change, expected behavior, primary goal, and guardrail metric. Example: simplifying the pricing-page hero will increase unique signup clicks without reducing plan-detail views.
-
2
Open Actions → Website → A/B testing
Select Create experiment. In Step 1 choose A/B changes or Split URL, name the experiment, set Eligible traffic included, and confirm the correct website source.
-
3
Configure page and audience targeting
In Step 2 set page rules such as full URL, path, hostname, query, or fragment. Add only necessary audience rules for device, browser, language, visitor type, referrer, UTM values, cookies, or viewport width.
-
4
Create variants and allocation
In Step 3 add at least one variation and set traffic weights. Save the draft, then use Open visual editor to select live elements; or add deliberate selector-based text, HTML, visibility, attribute, class, CSS, or JavaScript changes.
-
5
Set the decision metric
In Step 4 add Page view, Element click, Form submission, Custom event, or Revenue event goals. Configure the selector, URL, or event name, choose unique visitors or all events, and mark exactly one Primary goal.
-
6
Validate the launch-readiness panel
Confirm the website pixel was seen in the last 24 hours, variations are configured, and the primary goal is measurable. Preview control and variations on desktop and mobile, including forms, navigation, consent, analytics, and page speed.
-
7
Select Save & start experiment
Start only after QA. Assignments are deterministic and sticky. Once visitors are recorded, targeting, variants, and goals are locked to preserve the test.
-
8
Read results without stopping early
Open the experiment results to compare visitors, conversions, rates, uplift, revenue where relevant, and uncertainty. Cover normal weekly variation and document the business decision before completing the test.
Verify your work
- Launch readiness confirms a recent pixel, configured variations, and a measurable primary goal
- The experiment status changes from draft to active
- Control and variation exposures appear in Results
If something doesn’t look right
Save & start is disabled
Finish at least one variation and select a valid primary goal.
Pixel warning remains
Generate a real page visit, check consent and installation, then confirm an event appears in Data → Own Data → Pixel Tracking.
A visual change is absent
Verify the URL rule, CSS selector, page timing, and that another script is not replacing the same element.
Limitations and safeguards
Cross-device visitors can be assigned separately when they cannot be recognized consistently; external scripts and site releases can also affect a live experiment.
Important notes
Do not declare a winner from a small sample or a short-lived spike.