Google AdsTo test a Google Search landing page, create a custom experiment from an active campaign and change the destination while keeping the intended comparison settings consistent. Define the conversion outcome, allocation, budget and review date before launch. Compare qualified outcomes after allowing for conversion delay, rather than choosing a winner from clicks alone.
This guide prepares a landing-page experiment for an imaginary scheduling SaaS. Download the Search experiment worksheet. The rows are planning examples, not measured campaign results. You will finish with a hypothesis, a configuration record and a decision rule that someone else can review.
Define the result before changing the campaign
Write the business outcome first. Our example counts a trial as qualified when a company creates a workspace and invites a teammate. A page visit or an email submission alone does not satisfy that definition. Both experiment arms must use the same conversion definition and attribution settings.
Record the primary metric and a practical constraint. The primary comparison could be cost per qualified trial; the constraint could be a maximum approved test spend. Specify who decides whether the evidence is sufficient. A smaller cost per click is interesting, but it does not answer the qualified-trial question by itself.
If your account only measures form submissions, complete the qualified-lead conversion workflow first. Changing measurement halfway through a landing-page experiment makes the comparison harder to interpret.
Freeze one difference between the two arms
For this example, the base campaign sends people to the general scheduling page. The experiment sends them to a page explaining restaurant scheduling. Keep the offer, keyword selection, ad message, conversion actions and other intended settings consistent. If the restaurant page also offers a discount, you are testing the combined page and offer.
| Item | Base | Experiment |
|---|---|---|
| Destination | General scheduling page | Restaurant scheduling page |
| Offer | Same trial | Same trial |
| Primary outcome | Qualified trial | Qualified trial |
| Intended difference | Page experience | Page experience |
Save dated page versions and destination URLs. Open both on a phone, complete the same test journey and check tracking before scheduling anything. Use the Google Ads preflight to keep URL, copy and measurement checks together.
Create the experiment with its allocation recorded
In Google Ads, open Campaigns, then Experiments, and create a Custom experiment for Search. Select the active base campaign, give the experiment a recognizable name, and configure the treatment. Google's setup instructions describe the goals, traffic split, dates and advanced allocation controls. Availability depends on the campaign configuration; check restrictions before proceeding.
Search experiments offer cookie-based or search-based allocation. The choice affects whether a user remains in one arm or can encounter different arms across searches. Record the selected method in the worksheet rather than leaving it implicit. A 50/50 split is a reasonable starting proposal for this example, subject to the campaign owner's approval.
Review the actual campaign, budget and start date before saving the scheduled experiment. This action can send paid traffic when the schedule begins. The worksheet does not connect to an account or make that decision for you.
Keep a change log while the test runs
Capture the configuration at launch and note any necessary interventions. Record outages, tracking failures, changed offers and page edits with their timestamps. Do not silently repair a treatment page and then describe its entire history as one unchanged experience.
Google's experiments overview explains that allocation does not guarantee identical realized spend or impressions. A spending difference alone is therefore not proof that the test is broken. Investigate eligibility, delivery and the actual settings before drawing that conclusion.
Give conversions time to arrive. The conversion-lag review helps separate immature reporting from a real difference. Choose review dates in advance; repeated daily winner declarations encourage decisions based on temporary fluctuations.
Close with a decision and its limits
Export the experiment report using a consistent date range. Keep primary outcomes, spend, reported uncertainty and known incidents together. Compare the result with the hypothesis written before launch. If volume is too low or tracking changed, “inconclusive” is a useful outcome.
For our imaginary team, the final note might read: “The restaurant page produced a promising qualified-trial result, but the last reporting days are incomplete. Keep the original campaign and review again after the recorded conversion window.” This is an example of a decision format, not a claimed test result.
Before applying a treatment, inspect exactly which settings will carry over and retain the previous version. A useful handoff includes the report, changed destination, approver and review date. The next experiment should start from that documented state.
Sources checked October 15, 2026. This is a source-reviewed planning workflow; no advertising spend or live experiment was run for this article.