Pages > Experiment
Experiments
Run A/B tests across page variants to find which version converts best.
Overview
Experiments let you run a controlled A/B test using two or more variations within a page variant. SuperFunnel splits traffic between them, measures performance against your chosen metric, and tracks results until your completion criteria are met.
A quick distinction: variants split your audiences (e.g. English visitors vs. Spanish visitors). Variations split traffic within a variant — the same audience sees different designs to find which performs best. Experiments test variations, not variants.
Before running an experiment, you need a variant to run it on. If you haven't set one up yet, see Page Variants.
Choosing your metrics
Pick a primary metric that matches your actual goal — Conversion Rate if you care about efficiency, Conversions if you care about total volume.
What this guide covers
- Starting an experiment
- Configuring the experiment settings
- Building each variation
- Previewing variations before collecting data
- Viewing results in the Experiments Hub
Running an Experiment
Open your page
Go to Pages from the left sidebar and click the page where you want to run an experiment.
Start the experiment
Find the variant you want to test and click Experiment. The Start Experiment panel will open.
Configure the experiment
Fill in the experiment settings on the left-hand panel:
Experiment Name — give it a descriptive name so you can identify it later. For example: "Headline Test – Urgency" or "Mobile CTA Optimisation".
Name experiments with a date
Include the testing period or start date. For example "Headline Test – Jan 2026". This makes it easy to distinguish similar experiments when reviewing results later.
Primary Metric — the metric the experiment is optimising for. Options include:
- Conversion Rate
- Conversions
- Button Clicks
- Form Submissions
Secondary Metrics (optional) — additional metrics to track alongside the primary. Useful for spotting trade-offs or monitoring other signals.
Completion Criteria — when the experiment should end. Options include:
- Manual — you decide when to stop and review results
- Statistical confidence threshold — experiment ends when 95% confidence is reached; you can also set a minimum number of visitors per variant and a minimum duration in days
- Fixed duration — set a number of days until the experiment ends
- Visitor count — set a total number of visitors until the experiment ends
Add variations
The control variation (your original page design) is included by default.
4.a. Click + Add Variation to add each additional variation (treatment) you want to test. For each variation, set a description of what is being tested and its traffic allocation percentage.
4.b. All percentages across the control and all treatments must add up to 100%.
Let it go
Deleting a variation after an experiment has started can affect test results. Avoid removing variations once the experiment is running.
Start the experiment
Review your settings and click Start Experiment. Traffic will begin splitting according to your allocation and results will accumulate in Variant Analytics.
Build each variation
After starting the experiment, click Generate on each variation to open the page editor and define what that variation looks like. You will be prompted to choose a starting point:
- Start from control — begins with a copy of the control variation
- Start fresh — opens a blank canvas
- Copy from another page — pulls in the design from a different page or variant
Enter a prompt describing what should be different in this variation — for example: "Make this a dark mode version" or "Replace the hero headline with urgency-focused copy". Click Continue to send the prompt to the Agent, or Continue without prompt to open the editor directly.
Wait for the Agent to finish its changes, then make any additional manual edits needed. When satisfied, save and return to the experiment.
Build your variation first as an inactive variant
Before starting the experiment, create the variation by first adding a new inactive variant to the page, then use Copy from another page when generating the variation to pull that design in. This way you can refine the design fully before it goes into the experiment without rushing edits after the fact. Once the variation is generated, you can delete the inactive variant or leave it inactive for use in future experiments.
Switch variations in the editor
To reference another variation while editing, use the dropdown menu in the top-left corner of the editor to switch between variations without leaving the editor.
Preview all variations in production
Now that your experiment is live, load both the control and each treatment variation at their production URLs to confirm they look and behave as expected. Check layout, copy, forms, popups, hyperlinks, redirects and any interactive elements.
If a variation has unintended issues — broken layout, wrong copy, or missing elements — go back and make corrections before traffic accumulates on a flawed version. Results from a broken variation are not meaningful and can skew the experiment.
Viewing Experiments
Click Experiments in the left sidebar to open the Experiments Hub — a central view of all your ongoing and completed experiments.
Each row in the hub represents one experiment and shows:
- Experiment name — with the associated page name listed below it. Click the page name to go directly to the page view where you can see variants and traffic options.
- Status — Active or Completed
- Duration — how many days the experiment has been running
- Confidence — the current statistical confidence percentage
- Variations — the number of variations included in the experiment
Click any row to expand it and see a breakdown of each variant's performance.
Click the vertical ellipsis (⋮) on the right side of a row to access additional options:
- View page — navigate to the page associated with this experiment
- Edit Experiment — update the experiment name, metrics, or completion criteria
- Complete Experiment — manually end the experiment and lock in the results
What's Next
Once the experiment is running, monitor results in the Experiments Hub and wait for your completion criteria to be met before drawing conclusions.