Define the hypothesis and the real conversion
Start with a falsifiable statement: replacing a generic service headline with a search-intent-specific promise will increase qualified consultations without lowering the percentage of leads accepted by sales. Name the audience, change, expected mechanism and measurable result before opening an experiment tool.
Choose a primary outcome close to business value. For a service business that may be sales-qualified leads per eligible ad click, with raw submissions, appointment bookings and bounce rate as diagnostic metrics. If the sales cycle is long, define a leading indicator and a later CRM reconciliation date rather than claiming early forms prove success.
Audit the traffic before comparing pages
Export the last four to eight weeks of search terms, device mix, location, match type, conversion actions and landing-page paths. Identify whether traffic is homogeneous enough to compare. If the proposed change only helps emergency searches, do not dilute the test with broad informational queries.
Keep campaign targeting, bidding goals, budgets and creative as stable as practicable while the page is tested. Document any changes outside your control, such as seasonality, a competitor promotion or a form integration incident. A test is not interpretable when several meaningful variables change at once.
Select an experiment method that fits the implementation
For a true page-only test, use a reliable website-side A/B assignment that persists per visitor and records the variant in analytics and the CRM. Verify both pages share tracking definitions and neither version accidentally loads different consent or form code. A redirect or split test should preserve campaign identifiers and avoid duplicate conversion events.
Google Ads custom campaign experiments are designed to compare changes to campaign configurations and have eligibility and split rules. They are not automatically the right tool for every landing-page-only test. If using campaign experiments, follow Google's supported campaign types and avoid changing the base campaign mid-test.
Plan sample size and stopping rules
Estimate the baseline qualified conversion rate and the smallest improvement that would change a business decision. A page with a low baseline rate needs substantially more visits than a high-volume consumer checkout to detect a modest lift. Use a sample-size calculator or analyst-reviewed plan before launch.
Write a minimum observation window that covers weekday and weekend patterns, a minimum eligible-visit target, and a maximum run date. Do not stop because one variant is ahead on day three. Avoid running many simultaneous variations against sparse traffic; a sequential qualitative usability test may be more responsible.
Quality-assure the full journey
On desktop and mobile, click a tagged test ad or safe preview link, confirm the assigned page, complete the form, verify the thank-you state and inspect the CRM record. Check that UTM parameters, Google click identifiers where legitimately available, variant ID and qualification answers survive the journey.
Test failure cases as well: blocked cookies, validation errors, phone calls, returning visitors, duplicate submissions and abandoned forms. Record the exact timestamp and expected event for each test. If a conversion fires on a button click rather than a confirmed submission, repair measurement before starting the experiment.
Analyze outcomes and make a decision
Report eligible clicks, page sessions, valid submissions, qualified leads, booked calls, customers and cost at each stage by variant. Check whether differences are concentrated on mobile or a particular query theme. When CRM data matures, revisit the result rather than freezing the conclusion at the form-submit stage.
An example: variant B raises forms by 18% but the share accepted by sales falls enough that qualified leads per click remain unchanged. That is not a business win. If sample size is inadequate, label the result inconclusive, document what was learned and design a larger or more targeted follow-up.
Operationalize the winning change
Promote a winner only after checking accessibility, page speed, consent, SEO canonicalization and ongoing lead delivery. Save screenshots and a change log, and monitor the winning version for at least one full sales-response cycle.
Use the Google Ads experiments guide for campaign-level test design, the website message-match guide for relevance and the funnel calculator for scenario planning. The output should be a reusable decision record, not a collection of disconnected screenshots.