# Landing page A/B testing: a launch and decision checklist

Plan a landing page A/B test around one change, confirm that traffic reaches both versions, and decide when the result is ready to act on.

Canonical page: https://cpcinsider.com/insights/landing-page-ab-testing-checklist/
Author: Vladlens Kecko
Published: 2026-11-14
Updated: 2026-11-14

A landing page A/B test answers a narrow question: does a defined change improve an outcome for eligible visitors? Sending different campaigns to two pages answers a different question, because the visitors may differ before they see either version. Use this checklist for a website experiment with randomized assignment. For tests of campaign settings, see the [Google Ads experiments guide](https://cpcinsider.com/insights/google-ads-experiments/).

## 1. Write the decision rule before launching

Keep one meaningful difference, such as the number of required form fields. Define who is eligible, what gets assigned, and how repeat visits keep the same assignment. Use one primary outcome, such as the proportion of assigned visitors who produce a CRM-confirmed qualified enquiry within an agreed follow-up window. Define a guardrail, such as error rate, and decide what worsening would block rollout. Set the minimum business-relevant effect, target sample and analysis point with your experiment tool or analyst before launch. If ordinary traffic cannot support the required sample, make a diagnostic or usability improvement and document the limited evidence.

## 2. Verify the entire path in both versions

Check the assignment, rendered variant, form response, recorded event and matching CRM record. Exclude controlled test traffic according to your measurement plan. Confirm the denominator uses the original assignment consistently, and log failures without dropping affected visitors. During the run, look for a meaningful mismatch between the configured split and observed assignments. Investigate routing, assignment and event collection if your tool flags an imbalance before treating any variant as a winner. [Microsoft’s experimentation team explains assignment checks](https://www.microsoft.com/en-us/research/articles/patterns-of-trustworthy-experimentation-pre-experiment-stage/) and [checks during a test](https://www.microsoft.com/en-us/research/group/experimentation-platform-exp/articles/patterns-of-trustworthy-experimentation-during-experiment-stage/).

## 3. Review mature outcomes and make the decision

| Check | Decision |
| --- | --- |
| Allocation or measurement failed | Diagnose and rerun after fixing the problem. |
| Sample is immature or uncertainty is too wide | Record the result as inconclusive and follow the predefined rule for extending or closing. |
| Meaningful improvement with acceptable guardrails | Roll out in stages and monitor the same outcome. |

In a hypothetical test, shorter forms produce more submissions but the qualified enquiry rate remains too uncertain to distinguish the variants. The correct decision is to hold the rollout and record the uncertainty. Treat observed lift, uncertainty, and business value as separate inputs. Adjust for multiple comparisons if your design tests multiple versions or outcomes.

Use the [paid traffic landing page guide](https://cpcinsider.com/insights/landing-page-optimization-paid-ads/) to find a testable issue and our [tracking and CRM service](https://cpcinsider.com/services/tracking-attribution-crm/) to connect outcomes. [Book a free 30-minute founder-led growth review](https://cpcinsider.com/#audit) to receive a short written plan with three priorities.