How to Run an A/B Test on WordPress
Run a WordPress A/B test with a control, one variant, a real conversion goal and prelaunch checks. Includes the current Nelio workflow and common setup issues.
To run an A/B test on WordPress, use an experiment tool to assign eligible visitors to an original page and a variation, measure the same outcome for both and evaluate the result using a planned stopping rule. Duplicating a page and comparing two different weeks is not the same experiment.
The walkthrough below uses Nelio A/B Testing because its current documentation describes a WordPress dashboard workflow. Niwa can help with supported page edits and customer assistance; the experiment platform handles variant allocation and measurement. You do not need to replace your WordPress website to add that assistance.
Prepare the test before installing or starting anything
Choose one working page, one meaningful change and a primary outcome. For example, compare the existing service-page introduction with a version that explains what a quote request includes. Keep the service, price, form and campaign consistent.
Check the tool's current plan, supported editor and usage allowance against the eligible traffic you expect. A plugin download does not establish that every experiment or traffic level is available without a separate service plan.
Back up the relevant content and verify the journey on staging. Confirm that an enquiry can be submitted and reaches the team. Record your current page and tracking configuration so you can check them after the experiment ends.
Create a page experiment in Nelio
The official creation guide provides this sequence: open Nelio A/B Testing, choose Tests, select Add Test and select the appropriate type. For this example, use a Page Test and give it a name describing the question.
Set the scope to the page you intend to evaluate. Choose the current version as the control and add one variation. Open the variation in the supported WordPress editor and make the planned change.
Check that only the intended content differs. If you also change the form, images and offer, the experiment evaluates the combined package. It cannot tell you that the introduction alone produced the result.
Record how the tool maintains a visitor's version during the journey. Avoid a route where people see one promise on the landing page and a conflicting promise after the next click.
Configure an outcome that represents completion
Choose the goal and conversion action available for your journey. Use a verified purchase action for a store or a success event for an enquiry when the integration supports it. If you use a confirmation-page view, ensure people reach that page only after the request succeeds and account for repeat visits.
A click can be a secondary diagnostic measure. It should not stand in for a saved appointment or delivered enquiry when those are the outcomes you need.
Track a guardrail outside the headline number, such as enquiry qualification, failed submissions or refunded orders. The lead-capture guide explains why capturing contact information and obtaining a useful lead are different parts of the workflow.
Check delivery, caching and tracking
Before exposing real visitors, view both versions with the tool's supported staging or debug features. Test desktop and phone layouts, navigation, the success state and the experiment's recorded outcome.
Nelio's compatibility documentation identifies conflicts involving script optimization, caching, security rules and URL-parameter cleanup. Use the documented exclusions needed for your installation, then clear the relevant caches and recheck. Avoid disabling every performance or security control to make a test run.
If one version never appears, inspect scope, allocation and caching. If the page changes but conversions do not record, inspect the goal and success event. If the variant breaks, stop and repair the technical problem rather than interpreting its failure as customer preference.
Configure any analytics and recording collection for the site's applicable privacy and consent requirements. A blocked tracker should not be treated as permission to override a customer's choice.
Run the comparison and make a decision
Launch only after both versions and the outcome check work. Monitor technical health during the test, but follow the experiment's statistical method and planned decision rule. A conventional fixed-sample test should not end simply because one version moves ahead early.
Use the actual eligible traffic and expected outcome rate to plan the sample. Include relevant business cycles. There is no universal instruction to run every WordPress test for seven days or stop after a set number of clicks.
Review the outcome, uncertainty and guardrails together. If the result is inconclusive, preserve that finding. If you select a version, apply it through the tool's supported process, stop the experiment, clear necessary caches and verify that customers now see the intended page.
Fit Niwa into the tested journey
Configure Niwa with the approved service information so the assistant answers the same offer shown on both versions. Keep that assistance consistent while testing the introduction. Use authorized WordPress operations for supported owner edits and verify the saved result.
Niwa Business is USD 150 for a custom WordPress website or migration, hosting, maintenance and AI adaptation. Existing WordPress owners can use the standalone plugin. Contact Niwa about the customer journey you want to support, with the page, intended action and experiment hypothesis.
