How do I test two versions of a page?
Short answer
Build two versions of the page that differ in exactly one meaningful element, split traffic evenly between them, and let each collect enough submissions to be compared fairly. Track which version produces more booked jobs, not just more form fills, before declaring a winner and retiring the loser.
Before you start
- A live landing page with steady traffic.
- A clear conversion goal, such as form submits or calls.
- A hypothesis about what change might improve results.
- Enough traffic volume to reach a fair sample size.
Step by step
- 1
Pick one variable to test
Choose a single element like the headline, form length, or main image, not several at once.
Changing multiple things at once makes it impossible to know what worked.
- 2
Build the second version
Duplicate the page and change only the one variable you decided to test.
Keeping everything else identical isolates the true effect of the change.
- 3
Split traffic evenly
Route roughly half of visitors to each version using an even random split.
An uneven split biases results toward whichever version got more traffic.
- 4
Define your primary metric upfront
Decide whether you are judging by form submits, calls, or booked jobs before the test starts.
Choosing the metric after seeing results invites bias toward whichever looks better.
- 5
Let the test run to a fair sample size
Wait until each version has enough visits and conversions to trust the comparison.
Ending too early often just measures random noise, not a real difference.
- 6
Track outcomes past the form
Follow leads from each version into the pipeline to see which produced more booked jobs.
A version with more submissions but lower booking quality is not actually winning.
- 7
Declare a winner and retire the other
Once the data is clear, keep the winning version live and remove the losing one.
Running both indefinitely just splits traffic without improving results.
- 8
Document the result
Record what changed and what happened so future tests build on what you learned.
A written history keeps you from retesting the same idea twice.
What good looks like
- You know which page version actually performs better.
- Decisions are based on booked jobs, not just clicks.
- Future page changes build on documented results.
- Traffic is not wasted on an underperforming version indefinitely.
Common mistakes
The ways this goes wrong in insights, and what to do instead.
Testing too many changes at once›
The headline, image, and form all change together and the result is unreadable.
Do this instead: Isolate one variable per test so the cause of any change is clear.
Calling the test early›
A slight early lead gets declared a winner before enough data has come in.
Do this instead: Wait for a meaningful sample size before ending the test.
Judging by submissions alone›
One version collects more forms but those leads rarely book real jobs.
Do this instead: Follow leads into the pipeline and compare booked job outcomes.
Never retiring the loser›
Both versions stay live forever, diluting traffic and confusing reporting.
Do this instead: Commit to removing the underperforming version once the test concludes.
Frequently asked
6 questions about test two versions of a pageRelated guides
Pick the next thing to set up, or back up a step if something here assumed work you have not done yet.
Do this next
The natural follow-on tasks once this one is working.
People usually get here from
Guides that lead into this one, in case you skipped a step.
More in Insights
Nearby tasks that share the same setup and vocabulary.
Was this guide helpful?
Tell us what worked and what didn't. It shapes what we rewrite next.
You didn't get into business to be in the tech business.
Managed Tech clients hand this to us. We build it, test it, maintain it, and add to it as the business changes, so nobody on your team has to become the person who knows how the system works.