From install to a ready-to-launch first test.
Install the snippet, create a test, make one change, preview it, and let the results come in. Here is how the whole workflow fits together.
Five steps. No ceremony.
Install once, write the test, edit the page, measure the result, and record what you learned.
Add the snippet to your site.
Copy one line of JavaScript and paste it before the closing </head> tag. It works with static HTML, WordPress, Webflow, Squarespace, Shopify, and most sites that accept a script tag.
The snippet is under 8KB and loads asynchronously. Once SplitPea detects it, you are ready to create a test.
data-site="your-site-id"></script>
Set up an experiment.
Give it a name, write down what you think will happen, pick the page, and choose the conversion goal.
The hypothesis is optional, but useful. “A shorter headline will get more signups” takes ten seconds and gives the result context later.
Make your changes visually.
The visual editor opens your actual website in the browser. Click the element you want to change and type the new version.
You can change headlines, button text, paragraph copy, images, and simple styling. CSS selectors are available when you want more precise targeting.
Watch the numbers come in.
SplitPea splits traffic evenly between control and the variant. Returning visitors keep seeing the same version.
The experiment page shows visitors, conversions, conversion rate, lift, and confidence. SplitPea tells you when the test needs more time.
Call the winner and record the decision.
When the result reaches SplitPea’s decision threshold, it shows a recommendation and the supporting numbers. End the test when you have enough data, apply the change yourself, and record the decision.
If neither version wins, the result is marked inconclusive. That is useful too, and it stays in the decision history.
That's the whole process.
Install once. Create experiments whenever you want to test something. Read the result. Keep a record. Your site gets a little smarter each time.
The bits people usually ask.
How long does an experiment take?
It depends on traffic and the size of the difference. A site with a few hundred visitors a day may see a result in a week or two. Lower-traffic sites take longer.
What if I do not have much traffic?
Test bigger changes and be patient. SplitPea keeps “not enough data yet” visible instead of forcing an early call.
Can I test more than two versions?
SplitPea currently supports one control and one variant. Simpler tests are easier to interpret, especially on smaller sites.
What happens to the losing version?
Nothing automatic. When the test ends, visitors see the original page. You decide when and how to apply the winner permanently.
Do I need to edit my site to apply the winner?
Yes. SplitPea tests the change; your normal site editor or codebase is where you make it permanent.
What does 90% confidence mean?
A 90% confidence score means SplitPea estimates a 90% probability that one version is better than the other under its model. It is evidence, not a guarantee, and small differences may still be inconclusive.
Start with one clear question.
Write a hypothesis, draft one variant, and decide with evidence.