After running and analyzing over 1,000 A/B tests across eCommerce, SaaS, and lead-gen businesses, clear patterns emerge. Some ideas work 8 out of 10 times. Others fail 9 out of 10 times. This guide is the distilled wisdom from a thousand experiments: what actually works, what’s a waste of traffic, and why.
The Foundation: What You Need to Know
1 in 3 tests wins. This is the industry average. If your win rate is 50%+, you’re either lucky, running huge-effect tests, or not running truly novel tests. If it’s under 20%, your hypothesis quality is weak or your sample size is too small.
Data beats opinions by 2x. Tests grounded in user research, heatmaps, or session recordings win at ~65%. Opinion-based tests win at ~35%. The difference: informed hypotheses are more specific and more grounded in real behavior.
Traffic is your limiting factor. With 10k monthly visitors, you run 1 solid test/month. With 100k/month, you can run 5. Don’t speed up testing to 2/week if your traffic doesn’t support it. You’ll get noisy results and learn nothing.
What Consistently Works (Win Rate 70%+)
1. Reducing Friction
Friction = any step between intent and action.
Form field reduction: Removing one field typically lifts conversion 10–15%.
- 3-field form beats 7-field form: 85% win rate
- Removing “Company” field from signup: 75% win rate
- Removing “Phone number” when unnecessary: 70% win rate
Why it works: Every field is a barrier. If you don’t need it, remove it.
Guest checkout: Forcing account creation is a tax on conversion.
- Guest checkout enabled: 75% win rate, average 15% CVR lift
- Optional account creation after purchase: 80% win rate
- Why: Removing a decision step (account vs guest) accelerates purchase
Navigation simplification:
- Removing top nav from landing page: 70% win rate, average +12% CVR
- Reducing menu items from 8 to 4: 60% win rate
- Why: Fewer choices = fewer distractions = more focus on primary CTA
Page load speed: Every 1 second of load time costs ~5–7% conversion.
- Optimizing from 4s to 2s load: 90% win rate
- Why: Users abandon slow pages before even seeing the offer
2. Social Proof Near the CTA
Social proof works because it’s behavioral anchoring: “If others are buying, it must be safe.”
Product reviews near Add to Cart:
- Star rating + review count adjacent to ATC: 85% win rate, average +18% CVR
- User-generated photo with review: 80% win rate
- Why: It’s the last objection before purchase. Solve it there.
Testimonials on pricing pages:
- Customer testimonial emphasizing value/ROI: 70% win rate
- “Person X from Company Y saved $Xk” specific testimonial: 75% win rate
- Why: Pricing pages are high-friction. Social proof reduces skepticism.
Trust badges on checkout:
- SSL badge + money-back guarantee + return policy: 65% win rate
- Known payment logos (Stripe, PayPal): 60% win rate
- Why: Checkout is the moment of highest skepticism. Reduce risk.
3. Clarity Over Cleverness
Users don’t remember clever copy; they remember clear benefits.
Benefit-focused headlines beat clever ones: 80% win rate
- “Save 20% on your first order” beats “Get more for less”
- “Free shipping on orders over $50” beats “We ship far, you save more”
- Why: Clarity removes guesswork. The reader knows what you’re offering.
Descriptive CTAs beat generic ones: 90% win rate
- “Get my free audit” beats “Submit”
- “Start my free trial” beats “Sign up”
- “See pricing” beats “Learn more”
- Why: Specificity sets expectations and reduces friction.
Showing price early beats hiding it: 75% win rate
- Price visible on product page: higher conversion than “Add to cart to see price”
- Annual pricing shown first: higher AOV than “Select plan first, then see price”
- Why: Transparency builds trust. Price hiding creates suspicion.
4. Mobile-Specific Optimizations
Mobile is now 60–70% of traffic for most sites. Mobile-specific tests often win big.
Sticky mobile CTA:
- Persistent bottom button on mobile: 70% win rate, average +20% mobile CVR
- Why: On mobile, users scroll past your CTA. Stick it to the bottom.
Express checkout on mobile:
- One-click Apple Pay / Google Pay: 80% win rate, average +25% mobile CVR
- Why: Mobile conversion is all about friction. Express checkout eliminates it.
Simplified mobile navigation:
- Hamburger menu instead of top nav: 65% win rate
- Mobile-optimized header (larger text, tap-friendly): 70% win rate
- Why: Mobile screens are small. Simplify.
5. Visual Hierarchy
Your eye should move from headline → value prop → CTA → Done.
Making the primary CTA more prominent: 65% win rate
- Larger button: +10% CVR
- Contrasting color: +8% CVR (if changed from subtle to visible)
- Above the fold: 15% CVR lift
- Why: Attention is limited. Direct it.
Removing competing CTAs from the same section: 70% win rate
- One CTA per section: better than three
- Why: Too many choices = analysis paralysis. Pick one.
What Rarely Works (Win Rate under 25%)
1. Button Color Changes
The most famous CRO “test” is also the least impactful.
- Red button vs green button: 15% win rate (barely above random)
- When it does win, lift is typically under 3%
- Exception: changing from a color that blends into the background to one that stands out (still works)
Why: Button color is a too-small effect for most traffic levels. You need massive traffic to detect a statistically significant 2–3% difference.
What to do instead: If you’re testing button color, you’re out of bigger ideas. Focus on friction, clarity, or social proof first.
2. Copywriting Micro-Changes
- Changing a single word in the headline: 20% win rate
- Removing a comma or changing punctuation: waste of traffic
- Capitalization (Title Case vs title case): no significant effect
Exception: Different value propositions DO work. If you’re testing “Save 20% today” vs “Free shipping on orders over $50,” that’s not micro-copy; that’s a different offer. Test it.
3. Layout Rearrangement Without Purpose
- Swapping left/right alignment: 25% win rate (usually inconclusive)
- Moving sections down one section: minimal effect
- Changing from 2 columns to 1 column with no other changes: inconclusive
Exception: Moving key content above the fold works consistently. “Removing navigation clutter and putting your USP higher” works. But “swapping the order of two equal sections” doesn’t.
4. Adding More Content
- Longer product descriptions (500+ words vs 200 words): 30% win rate
- More images without purpose: inconclusive, often hurts mobile performance
- More testimonials (3 vs 10): diminishing returns after 3–4
Exception: Adding missing information works. “What’s the size?” “What’s the shipping time?” “How long does the trial last?” If you’re answering a real objection, test it. If you’re just adding more words, it hurts.
Tests That Work by Page Type
Product Pages
| Test | Win Rate | Avg Lift |
|---|---|---|
| Better/larger product photos | 80% | 15% lift |
| Video product demo | 60% | +8% (if video is high-quality) |
| Star rating near ATC | 85% | +18% |
| Simplified “Add to Cart” form | 75% | +12% |
| “Only 2 left in stock” urgency | 70% | +10% |
| Remove product tabs, show key info inline | 65% | +8% |
Checkout Pages
| Test | Win Rate | Avg Lift |
|---|---|---|
| Reduce form fields | 85% | +12% |
| Progress indicator (step 1 of 3) | 70% | +8% |
| Guest checkout enabled | 75% | 15% lift |
| Security/trust badges on final step | 65% | +5% |
| Express payment (Apple Pay, PayPal) | 80% | +20% |
| “Order summary” visible on right | 60% | +6% |
Landing Pages
| Test | Win Rate | Avg Lift |
|---|---|---|
| Message match (ad headline = page headline) | 90% | +25% |
| Remove top navigation | 70% | +12% |
| High-quality video above fold | 55% | +7% (varies wildly) |
| Single-column form | 75% | +10% |
| Removing distracting elements | 70% | 15% lift |
| Social proof (logos, testimonials) | 75% | +12% |
SaaS Pricing Pages
| Test | Win Rate | Avg Lift |
|---|---|---|
| ”Most Popular” badge on middle tier | 80% | +20% plan selection |
| Annual billing as default | 85% | +25% ACV |
| FAQ section below pricing | 70% | +8% CVR + fewer support questions |
| Feature comparison table | 75% | +10% conversion |
| Removing one tier (simplification) | 60% | varies |
Meta-Lessons: How to Run Tests That Win
Lesson 1: Velocity Compounds
Running 2–3 tests/month (not rushing, being rigorous) = 24–36 tests/year = ~8–12 winners/year = compounding 12–18% improvements.
Running 10 tests/month (rushing, underpowered) = 120 tests/year = inconclusive results = you learn nothing.
Lesson 2: Big Changes Beat Small Tweaks
With 10k/month visitors and 2% CVR (200 conversions/month), a 5% lift is 10 conversions. That takes 3+ months to detect statistically.
A 15% lift is 30 conversions. That takes 3–4 weeks.
If you have limited traffic, test big changes. Friction reduction, CTA clarity, social proof. Not button color.
Lesson 3: Document Losing Tests
A losing test that’s well-documented (“We tested X; it lost because Y”) informs future tests.
A winning test that’s undocumented teaches you nothing.
Always document: hypothesis, why you expected it to win, what actually happened, and the learning.
Lesson 4: Data > Opinions
Tests grounded in heatmaps, session recordings, customer interviews, and analytics win at 2x the rate of opinion-based tests.
Before you test an idea, ask:
- Where’s the data supporting this?
- What page behavior or user feedback suggested this?
- What metric will change and by how much?
Lesson 5: Context Is Everything
What wins for a B2B SaaS pricing page might lose for eCommerce. What wins for high-AOV products might lose for low-AOV.
Copy the patterns, not the exact tests. The pattern: “Social proof near CTA works.” The specific test: depends on your context.
Common Testing Mistakes
Mistake 1: Running too many tests in parallel Confounding variables. You can’t tell what caused the lift.
Mistake 2: Stopping tests too early “We have a winner in 1 week!” No. You have noise. Wait for statistical significance (usually 2–4 weeks).
Mistake 3: Not controlling for external factors Holiday, PR, competitor move, traffic source change. If something external changes, the test is compromised.
Mistake 4: Testing ideas without data “Let’s try urgency” without ever seeing rage clicks or abandonment signals. Low win rate.
Next Steps
- Look at your last 5 tests. Which won, which lost, why?
- Pull your heatmaps and session recordings. What friction do you see?
- Test one of the patterns above. Pick friction reduction, social proof, or clarity. Run it for 3–4 weeks.
- Measure and document. Win or lose, write down what you learned.
- Repeat. The compounding returns come from volume + learning.
The best CRO programs aren’t the ones with the highest win rate. They’re the ones that test the most, learn the fastest, and compound improvements over time. Start with patterns that win 70%+ of the time. Then optimize from there.