Five tests that moved real numbers
At Spring Education Group, I ran A/B tests across multiple brands and landing pages to identify what was hurting conversions and prove what would fix it.
How I run every test
I don't run tests on hunches. Every experiment follows the same process: it starts with a real problem and ends with data that either proves or challenges what I thought I knew.
Virtual Open House page
The original VOH page wasn't answering the question families had when they landed there: why should I show up to this? No clear value prop, no urgency, and the RSVP path was too easy to miss.
The redesign refocused the page around the benefits of attending, made the RSVP impossible to miss, and added social proof through faculty and student ambassador content.
Hypothesis: Updating the Virtual Open House landing page can improve visit-to-inquiry rate. The previous page didn't focus enough on the advantages of attending and was missing components that drive conversions.
| Sessions | Calls or Apps | CR % | CR% Change | Prob. to Be Best | Expected Loss | |
|---|---|---|---|---|---|---|
| Original | 589 | 39 | 6.62% | - | 1% | 62% |
| Variant 1 | 653 | 69 | 10.57% | +4% | 99% | 6% |
Variant 1 had a 99% probability of being the best performer. CR jumped from 6.62% to 10.57% - a 4% lift. The original had a 62% expected loss if kept. Test ran April 21 to May 5, 2023. Estimated financial impact: $182,891.
More RSVP clicks means more attendees, and attending a Virtual Open House has a direct correlation with enrolling at Laurel Springs.
Revised homepage
The original homepage was long, cluttered, and felt dated next to competitors. The new variant was shorter, more elevated, and more focused - less content competing for attention, cleaner first impression.
Hypothesis: Shortening and reformatting the homepage will be more on par with competitors and increase leads.
| Sessions | Calls or Apps | CR % | CR% Change | Prob. to Be Best | Expected Loss | |
|---|---|---|---|---|---|---|
| Original | 2,181 | 31 | 1.42% | - | 4% | 53.16% |
| Variant 1 | 2,325 | 49 | 2.11% | +48% | 96% | 9% |
The new homepage had a 96% probability of being the best performer. CR jumped from 1.42% to 2.11% - a 48% relative lift. Test ran March 21 to March 27, 2023. Estimated financial impact: $2,036,009.
Better design meant better first impressions. The variant brought in more leads and reduced bounce.
Merryhill School homepage
This test was run specifically for direct visitors to the Merryhill School homepage - people who already had some brand recognition. The hypothesis was that since they already knew Merryhill, the hero messaging could be more targeted and action-oriented rather than purely introductory.
The idea was to change the hero to direct traffic toward a specific callout - something like "Interested in learning more?" - messaging that meets a warmer audience where they are rather than treating them like first-time visitors.
Hypothesis: Since direct visitors already have brand recognition, shifting the hero copy to be more action-oriented and audience-specific will increase new leads and click-through on key CTAs.
Variation 1 - "Take The Next Step With Merryhill School." The winning variant.
| Variation | Visitors | New Lead | Hero Clicks | Find School Nearby | Visit Page: New Lead | Nav Find A School |
|---|---|---|---|---|---|---|
| Original | 1,538 | baseline | 0.013 | 0.025 | 0.014 | 0.33% |
| Variation 1 | 1,531 | +14.16% | +206.39% | -12.76% | +14.16% | +40.64% |
| Variation 2 | 1,614 | -61.02% | +95.35% | -22.26% | -61.02% | -4.71% |
Variation 1 was the clear winner - +14% new leads, +206% hero clicks, and +40% nav clicks compared to the original. Variation 2 actually hurt lead conversion significantly (-61%), showing that copy direction really matters. Test ran March 12 to April 13, 2026.
The winning headline - "Take The Next Step With Merryhill School" - worked because it mirrored the language prospective parents already use when they're close to making a decision. It met them where they were.
Chesterbrook Academy hero CTA
The original hero had a "Start Here" text link. Not a button, just text. It wasn't doing anything for conversions. The hypothesis was simple: replace it with a real CTA button that actually tells families what to do next.
Hypothesis: Adding a "Find Your Campus" CTA button to replace the non-CTA "Start Here" text in the hero will increase leads.
| Sessions | Conversions | CR % | CR% Change | Confidence | Power | |
|---|---|---|---|---|---|---|
| Original | 3,932 | 52 | 1.32% | - | - | - |
| Variant | 3,914 | 75 | 1.92% | +45% | - | - |
A single CTA button swap drove a 45% conversion rate lift, from 1.32% to 1.92%. Sometimes the biggest wins come from the simplest fixes. Test ran March 1 to March 16, 2023. Estimated financial impact: $1,983,962.
This one is a good reminder that it's not always about a full redesign. The page didn't need a new layout. It needed a button that actually worked like a button.
Laurel Springs Course Catalog CTA
Users browsing the course catalog are actively engaged. They're deep in the consideration phase, comparing courses, reading descriptions. But there was no way to inquire from within the page itself. The "Request Information" button existed in the nav, but nav CTAs don't convert the same way in-context CTAs do.
Adding a "Request Information" button directly on the catalog page removes that navigation barrier and meets users at the moment of highest intent.
Hypothesis: If we add a "Request Information" CTA button directly on the course catalog page, the catalog visit to inquiry submission rate will increase, because users actively engaged with course content currently have no visible path to inquire from within the page itself.
| Visitors | New Lead | Conversions/Visitor | Improvement | Statistical Significance | |
|---|---|---|---|---|---|
| Original | 878 | 34 | 0.039 | - | Baseline |
| Variation 1 | 854 | 55 | 0.064 | +66.31% | 8% |
Variation 1 showed a +66.31% improvement in new leads. Statistical significance was at 8% when the test concluded. Lower than ideal, but the directional signal was strong enough to inform the next iteration.
Sometimes a test ends before significance is reached. That doesn't make the data useless: a 66% lift in early results is still a signal worth acting on.