A/B Test 96: Trust-forward mortgage page with process steps increased CTR 99%

The question

When a visitor lands on a mortgage product page, what does it take to get them to take action?

Variant A contained many of the elements commonly found on mortgage pages: rate information, loan options, a mortgage calculator, and multiple contact paths. The page provided information and resources, but the team wanted to understand whether a different presentation of that information could generate more engagement.

This test asked whether a comprehensive redesign focused on messaging, page structure, CTAs, and visual presentation would meaningfully increase the rate at which visitors clicked through to begin the mortgage process.

The hypothesis

The existing page presented mortgage information in a largely informational format.

The hypothesis was that increasing the visibility of key actions, refining messaging, introducing additional trust-oriented content, and improving the overall presentation of the page would encourage more visitors to engage with mortgage-related actions.

The expectation was that a redesigned experience would generate higher click-through rates than the existing version.

What changed

This was a comprehensive redesign rather than an isolated variable test.

Multiple elements changed simultaneously, which means the experiment evaluates the redesigned experience as a whole rather than any individual component. The results show how visitors responded to the package of changes, not which specific change drove the outcome.

Here is what changed between variant A and variant B.

The results

Variant B won with 99% statistical confidence.

The result is clean.

Variant B nearly doubled click-through rate, increasing conversion rate from 12.00% to 23.92%.

With 334 total conversions across 1,841 users, the experiment generated substantial data volume. The result achieved 99% statistical confidence.

Verdict: Win.

One important note on interpretation: because multiple elements changed simultaneously, the experiment demonstrates that the redesigned experience outperformed the control. It does not identify which individual change was responsible for the observed lift.

Design questions this test raises

How did visitors respond to the revised CTA presentation? Variant B introduced different CTA language, additional CTA placements, and new opportunities for visitors to engage throughout the page. These changes occurred alongside a broader redesign, so the experiment does not identify the contribution of any individual CTA modification.

How did visitors respond to the revised presentation of mortgage information? Variant B introduced new informational content, reorganized existing content, and presented key information differently than the control. The redesigned experience generated substantially more engagement, though the experiment was not designed to evaluate the impact of any individual section independently.

What role did content organization play? The redesign changed how mortgage-related information was structured and presented throughout the page. New sections were introduced, existing sections were consolidated, and some content was removed entirely. Because these changes occurred simultaneously, the experiment cannot isolate the impact of any individual organizational decision.

How did visitors respond to the revised messaging? Variant B introduced different messaging throughout the page, including updated value statements and revised calls to action. These messaging changes occurred alongside visual and structural modifications, making it impossible to separate their individual contributions.

What role did the overall redesign package play? The redesign modified messaging, visual presentation, content organization, navigation, CTA strategy, and page structure simultaneously. The experiment demonstrates that the redesigned experience outperformed the control, but it does not identify which individual components contributed most to the observed lift.

Why it matters

Even without knowing which individual element drove the result, the data points to some consistent patterns across the changes that are worth examining for anyone working on credit union website design or community bank web design. Rather than analyzing each section separately, it’s more honest to look at the themes that run across the full redesign.

The page shifted from passive to active language. Across multiple sections, variant A used exploratory framing: “Mortgage Solution Center,” “View Rates,” “Let’s talk.” Variant B used forward-motion framing: “Get Started Today,” “Estimate your payment,” “Connect with a mortgage professional.” Whether this language shift drove the lift is unknown. What is consistent is that the shift appears throughout the redesign, not just in one section.

Trust was stated rather than assumed. Variant A implied its credibility through the existence of rates and loan options. Variant B named it: expert guidance, competitive rates, simple process. These aren’t unusual claims for a mortgage page. What changed is that they were stated explicitly rather than implied. This is a design observation, not a data finding from this test.

The process became more explicit. Variant B introduced a “Steps to Homeownership” section and added additional guidance throughout the page. Whether these changes contributed materially to the lift cannot be determined from this experiment, but they represent one of the most visible differences between the two experiences.

Decision points were reduced at the close. Variant A gave visitors three contact options at the end of the page. Variant B gave them one. A reasonable argument exists that reducing optionality at the decision point lowers friction, but this is one pattern among many that changed. The data supports the overall direction, not this mechanism specifically.

Visual coherence improved throughout. Moving from a red/blue mixed palette to a consistent dark blue and pink system created coherence that wasn’t present before. Whether this contributed to the lift is unknown. What the data shows is that the full package, including this change, performed dramatically better.

What this means for your site

A 12% product page CTR can look functional while leaving significant conversions unrealized. Variant A was not a broken page. It had a calculator, rate information, loan details, and multiple contact paths. The 99% lift over that baseline is a reminder that a page doing the basics can still fall far short of what’s possible.

Comprehensive redesigns can be hard to learn from, but they’re often necessary to move the needle. This test changed multiple areas simultaneously. That limits what we can conclude about any individual element, but it’s also part of why the lift is 99%. Incremental tests often produce incremental results. A redesign that rethinks the full page, from hero through closing CTA, creates more surface area for improvement.

Information alone may not move visitors on these pages. Visitors who land on a mortgage product page typically know they’re interested in a mortgage. This test is one data point suggesting that more information doesn’t necessarily move them forward, while a clearer path and visible trust signals may. This is not a universal principle, and one test is not enough to establish it. But the direction is consistent with the broader pattern in this redesign.

Consolidating the closing CTA to one action is worth testing. If your closing section currently asks visitors to choose between calling, visiting a branch, emailing, and using an online tool, narrowing to a single primary action may reduce the friction of deciding. This test changed much more than the closing section, so the data doesn’t prove this mechanism specifically, but it is a reasonable hypothesis to test in isolation.

For teams working in bank website design or credit union web design, the 99% confidence on this test makes it a strong signal that the direction of variant B is worth adopting, even without knowing the precise mechanism behind it.

Test 96 was designed internally and ran for 18 days in spring 2025 across 1,841 sessions. All statistical confidence figures use MetriFi’s standard methodology. A result at or above 95% confidence is classified as a win.