Most websites have a conversion problem they have already paid to create. The traffic is there. The ad spend is there. The issue is a headline, a button, a form, or a layout that was chosen by preference instead of by data.
A/B testing — also called split testing — runs two versions of a page element against each other simultaneously, measures which one converts better, and gives you a number to act on instead of an opinion. Google Optimize was sunsetted in 2023, but tools like VWO, Optimizely, Convert, and the free tier of AB Tasty have filled the gap. The testing infrastructure is not the hard part. Knowing which tests are worth running is.
These nine are. They cover the elements that move real conversion numbers — not the color of a button for its own sake, but the decisions that determine whether a visitor understands your offer, trusts it, and acts on it.
Quick reference — the 9 tests:
- Headline framing and value proposition
- Primary CTA button copy
- Form length and required fields
- Hero image versus explainer video
- Social proof placement and format
- Pricing or offer presentation
- Long-form versus short-form page
- Sticky versus in-line call to action
- Urgency elements and guarantees
1. Headline Framing and Value Proposition
The headline is the single highest-leverage element to test because it determines whether a visitor reads anything else on the page.
Most headlines fail in one of two ways: they describe the business ("Full-Service Digital Marketing Agency in Los Angeles") or they describe a feature ("AI-Powered Campaign Management"). Neither answers the question a visitor is actually asking, which is: what changes for me if I stay on this page?
The three framings worth testing against each other:
- Outcome-led: "More qualified leads. Tracked from click to close." — the visitor gets to the result immediately.
- Pain-point-led: "Tired of paying for traffic that doesn't convert?" — meets the visitor where they are emotionally.
- Process-led: "We build the marketing system. You work the leads." — explains the mechanism for visitors who want to know how.
Run the test long enough to reach statistical significance — most testing platforms show you a confidence percentage; wait for 95% before calling a winner. On moderate-traffic pages, that often takes two to four weeks.
Takeaway: Test your headline before you test anything else. It is the first decision every visitor makes.
2. Primary CTA Button Copy
Button copy that names the specific outcome a visitor gets — "Get My Free Audit" instead of "Submit" — consistently outperforms generic commands.
"Submit" describes what the visitor does. "Get My Free Strategy Call" describes what the visitor gets. The distinction matters because at the moment of conversion, the visitor is weighing effort against reward. Generic copy reminds them of the effort. Specific copy reminds them of the reward.
Three variables to test in button copy:
- First-person phrasing: "Start My Free Trial" vs. "Start Your Free Trial" — first-person signals the visitor is already imagining doing it.
- Specificity of the offer: "Book a Call" vs. "Book a 30-Minute Strategy Call" — specificity reduces uncertainty about what happens next.
- The implied cost: "Get Started" vs. "Get Started — No Credit Card Required" — removing a perceived barrier in the button itself can lift clicks.
Button copy tests are fast to set up and fast to read results from because the click is a single, measurable action. Start here if you want a quick win.
Takeaway: Rename every generic button on your site by what the visitor receives, not what they do.
3. Form Length and Required Fields
Every additional required field reduces the number of people who complete a form, so testing a shorter version is one of the fastest ways to lift completion rate.
The Baymard Institute has documented this extensively in e-commerce checkout research — unnecessary fields are a primary source of abandonment. The same principle applies to lead generation forms. Each required field is a decision point. Some visitors decide not to continue.
The test: build a version of your form that asks only for what you need to make first contact — typically name, email, and one qualifying question — and run it against your current form.
Two things to track:
- Form completion rate — this will almost always increase with fewer fields.
- Lead quality downstream — if a shorter form produces more completions but significantly worse qualified leads, the net outcome may not be worth it. Track this in your CRM.
For most service businesses, phone number is the most abandoned field. Test removing it from the form and collecting it on the confirmation page or during the intake call instead.
Takeaway: Track form completion and lead quality together — a shorter form is only a win if the downstream quality holds.
4. Hero Image Versus Explainer Video
A short explainer video in the hero section can communicate a complex value proposition in 60–90 seconds that would take a visitor five minutes to read — but it comes with real trade-offs worth testing directly.
The case for video: visitors who watch an explainer video tend to spend more time on the page and arrive at the CTA with more context. For services that are hard to explain in a headline, video can close the comprehension gap faster than copy.
The case for a static hero image: a video player adds page weight. Slow load times hurt both conversion and Core Web Vitals, which Google uses as a ranking signal. On mobile connections, a video that buffers is worse than a strong static image with tight copy.
What to actually test:
- Autoplay vs. click-to-play — autoplay can feel intrusive; click-to-play requires the visitor to opt in, but you can add a strong thumbnail to drive clicks.
- Video hero vs. image hero at matched page speed — compress both versions to comparable load times, then test. You want to measure message effectiveness, not load speed.
Takeaway: Test video against image after optimizing page speed on both — otherwise you are measuring load time, not message.
5. Social Proof Placement and Format
Social proof works harder when it appears near the conversion action, not buried in the footer — placement often matters more than the proof format itself.
Most sites put testimonials in a dedicated section at the bottom of the page, after the visitor has already decided whether to convert. That is not where doubt lives. Doubt lives right next to the CTA.
Three placement tests worth running:
- Below the hero CTA: A single strong testimonial — one sentence, a real name, a specific outcome — placed immediately below the primary button addresses hesitation at the moment it occurs.
- Near the pricing section: If your page has a pricing block, proof directly adjacent to it (a quote from a customer who got clear ROI, or a recognizable logo) addresses the "is this worth it?" objection where it peaks.
- Inline with objections: If you have a section that addresses "how long does this take?" or "what if it doesn't work?", a supporting testimonial in that section is more persuasive than one collected on a separate reviews page.
Format also matters. Star ratings without context are weaker than a named quote with a specific outcome. "5 stars" tells a visitor nothing. "We cut our cost per lead by a third in the first 90 days — James T., HVAC business owner" tells them something real.
Takeaway: Move your single strongest testimonial to directly below your primary CTA and measure the effect before redesigning your entire proof section.
6. Pricing or Offer Presentation
How you structure and display a pricing section affects both conversion rate and which option visitors choose — without the underlying numbers changing at all.
The structural variables worth testing:
- Order of tiers: Most pricing tables go low to high. Some audiences respond better to high-to-low because anchoring to a higher price makes the middle tier feel more reasonable. Test both directions.
- Which tier is highlighted: A visual indicator ("Most Popular" or "Best Value") on the tier you want visitors to choose is a standard test. It shifts selections. Whether it shifts them toward your most profitable offer is what you are measuring.
- Guarantee visibility near the price: A money-back or no-risk promise placed directly adjacent to the price — not three sections below it — addresses the hesitation that occurs at the price point. Test placing it in the pricing card vs. in a separate section.
For service businesses that do not publish pricing, the equivalent test is the offer framing: "Book a Free Audit" vs. "Book a 30-Minute Strategy Call" vs. "See If You Qualify" — each implies a different level of commitment and selectivity.
Takeaway: Test tier order and guarantee placement before testing the prices themselves — structure often moves the needle faster than the number.
7. Long-Form Versus Short-Form Page
Page length should match the commitment level of the offer — high-consideration services often convert better on longer pages that answer objections.
There is no universally correct page length. The right length is the one that answers every question a visitor needs answered before they will act — and no longer.
When long-form tends to win:
- High-consideration purchases or services — legal, financial, medical, or agency services where the visitor is committing real money or time. They have objections. A longer page that addresses them converts better because it replaces a sales call with written answers.
- Cold traffic — visitors who do not know your brand yet need more context to build enough trust to act.
When short-form tends to win:
- Retargeting or warm traffic — visitors who already know you, who have been to the site before, or who clicked a specific offer ad. They do not need to be convinced from scratch. A short page with a clear CTA converts faster.
- Low-friction offers — a free download, a newsletter signup, a free tool. The commitment is small; the page should match.
The test: build a stripped-down version of your page — hero, CTA, and three to five proof points — and run it against the full version against the same traffic source. Segment by traffic source if possible, because the answer may differ.
Takeaway: Run the long vs. short test separately for cold and warm traffic — the winning version is often different for each.
8. Sticky Versus In-Line Call to Action
A sticky CTA — a button or bar that stays fixed to the top or bottom of the screen as the user scrolls — keeps the conversion action visible throughout a long page. Whether it outperforms a standard in-line CTA depends on context.
When sticky tends to win:
- Long pages — once a visitor scrolls past your in-line CTA, it disappears. A sticky version keeps the option accessible without requiring them to scroll back up. On pages where the CTA appears in the top third, sticky is often the difference between a visitor who wants to convert and one who has to go find out how.
- Mobile — thumb-accessible sticky bars at the bottom of the screen match how people naturally hold a phone and interact with it.
When sticky can hurt:
- Short pages — if the page is one or two screens, a persistent sticky element can feel aggressive. The visitor has not had time to consider the offer before being followed by a button.
- High-intent audiences — visitors who arrived specifically to book or buy may find a sticky element distracting. Test it against your in-line version before assuming it helps.
Implementation note: a sticky CTA adds a small amount of complexity to track correctly. Make sure your conversion tracking is set up to distinguish clicks on the sticky element from clicks on the in-line version — otherwise you cannot read the test.
Takeaway: Test sticky CTAs primarily on long pages and mobile traffic; short pages with high-intent audiences often do not benefit from them.
9. Urgency Elements and Guarantees
Real urgency signals and risk-reversing guarantees address different visitor hesitations; testing which one moves your specific audience is more reliable than guessing.
These are two distinct levers that are often grouped together, but they work on different psychological mechanisms:
Urgency reduces delay. "Offer ends Friday" or "Only 3 spots available this month" tells the visitor that waiting has a cost. The critical constraint: urgency only works when it is real. A countdown timer that resets every time the page loads, or a "limited spots" claim on a page that has run for two years, erodes trust with the visitors most likely to notice — your highest-intent visitors. If you use urgency, enforce it.
Guarantees reduce risk. "No-obligation strategy call," "Cancel anytime," "If you're not satisfied after 30 days, we'll refund the difference" — these tell the visitor that acting does not trap them. They are especially effective for first-time buyers or visitors evaluating an unfamiliar service.
What to test:
- Guarantee present vs. absent near the CTA
- Urgency element present vs. absent (with real enforcement)
- Guarantee copy: "No obligation" vs. a specific refund/satisfaction policy
- Urgency format: deadline date vs. limited availability vs. price-change notice
Track which combination moves your specific audience — the mix that works for a warm retargeting audience is often different from what works for cold search traffic.
Takeaway: Use urgency only when you will actually enforce it, and place guarantees as close to the conversion action as possible where risk objections peak.
Before You Start: What Makes a Test Actually Valid
Running a test is easy. Running a test you can trust is harder. Three things determine whether your results mean anything:
Statistical significance. Most testing platforms calculate this for you. Wait for 95% confidence before calling a winner. Calling a test early because one variant is ahead after 200 visitors is how you make the wrong decision with high confidence.
Minimum sample size. There is no universal minimum, but a test based on fewer than a few hundred conversions per variant is unreliable. On low-traffic pages, you may need to run a test for four to eight weeks to collect enough data. Use a sample size calculator before you start.
One variable at a time. If you change the headline and the button and the image in the same test, you cannot know which change drove the result. Test one element per experiment. Multivariate testing is a separate methodology with its own sample-size requirements.
If you do not have the tracking infrastructure to measure conversions accurately, no A/B test will give you reliable data. Your website and tracking setup has to be solid before split testing is worth the effort.
Frequently Asked Questions
What should I A/B test first?
Start with the headline and value proposition on your highest-traffic page. It is the element with the broadest reach — every visitor sees it — and it determines whether anything else on the page gets read. Once you have a winning headline, move to CTA button copy, which is fast to test and fast to read.
How long should an A/B test run?
Long enough to reach statistical significance, which most testing platforms display as a confidence percentage. A common threshold is 95%. The time this takes depends on your traffic volume and conversion rate. On lower-traffic pages, four to eight weeks is often necessary. Ending a test early because one variant looks ahead after a short period is how tests produce wrong answers.
What are good things to A/B test?
The highest-impact tests tend to be elements visible above the fold before a visitor makes any decision: the headline, the primary CTA, and the hero visual. After those, social proof placement, form length, and page length are consistently worth testing across most business types.
How much traffic do you need to A/B test?
The answer depends on your baseline conversion rate and the size of the difference you are trying to detect. A page with a 2% conversion rate needs more visitors to detect a 0.5% improvement than a page with a 10% rate. Use a sample size calculator before starting — the inputs are your baseline conversion rate and the minimum improvement worth detecting. As a rough orientation, testing platforms generally need hundreds of conversions per variant to produce reliable data, not just hundreds of visitors.
Can I run multiple A/B tests at the same time?
You can, as long as the tests are on different pages or different, non-overlapping elements. Running two tests on the same page simultaneously creates interaction effects — a visitor who sees variant A of your headline and variant B of your CTA creates a combination that is not cleanly attributed to either test. Test one element per page at a time.
What is the difference between an A/B test and a multivariate test?
An A/B test compares two versions of a single element — one change, two variants. A multivariate test tests multiple elements simultaneously (headline, image, and button at the same time, for example) to find the best combination. Multivariate testing requires substantially more traffic to reach significance because there are more variant combinations to measure. For most businesses, A/B testing produces clearer, faster answers.
How do I know if my A/B test results are reliable?
Check three things: statistical confidence (95% is the standard threshold), sample size (enough conversions per variant, not just visitors), and test duration (at least one to two full business cycles to account for day-of-week variation). If any of these are insufficient, the result is a guess with a number attached to it.
What happens if neither variant wins?
A test with no statistically significant difference is still useful data — it means the element you tested is not the bottleneck. Move to the next test. The goal is to systematically identify which elements are limiting conversion, and ruling out variables is part of that process.
If your tracking is not set up to measure conversions accurately, none of these tests will produce reliable data. That is the infrastructure problem to solve first. If you want a review of what your current site and tracking setup can support — and where the highest-probability test opportunities are — book a strategy call.