Start with the tests that win most often for sites like yours.
In our archive that means product titles, price framing and the menu for ecommerce; CTA copy, plans and pricing display for SaaS; CTA copy, form fields and headlines for lead gen; and paywall design and trial length for apps.
All of these are cheap to build and sit on pages that most visitors see.
How we ranked them
We counted wins by element tag across the 1,824 tests in our database, split by business type, and ranked the elements by how many tests they won.
We left out the broad 'layout' and 'landing page design' tags, which won often but are redesigns, not first tests.
Each entry shows the win count and one real test from our research.
One honest caveat: people publish winners more than losers, so 1,516 of our 1,824 tests are marked won.
That's far above the roughly one in three ideas that improve their target metric in Microsoft's experiments, as Kohavi and colleagues reported.
Read the counts as 'tried often and often won', not as odds.
Our A/B test win rate page covers this in more depth.
The ranked list at a glance
Five tests per business type, ranked by win count within each type.
| Rank | Test | Change | Result | Who ran it |
|---|---|---|---|---|
| 1 | Product title and headline copy (ecommerce, won 121 of 132) | Benefit-led product title | +16.92% revenue per visitor | Dave Diederen |
| 2 | Price and discount framing (ecommerce, won 104 of 126) | 'Save X% per oz' size labels | +41% conversion rate | Blend Commerce |
| 3 | Navigation menu (ecommerce, won 102 of 128) | Mega menu with category descriptions | +$151,192/month | Oliver Kenyon |
| 4 | Upsells and add-ons (ecommerce, won 101 of 112) | One-click 'Frequently Bought Together' | +10.87% revenue per visitor | Alex Stoykov |
| 5 | Product image order (ecommerce, won 88 of 100) | Swapped first two gallery photos | £3.32 to £3.60 revenue per visitor | Elevate A/B Testing |
| 6 | CTA button copy (saas, won 19 of 23) | 'Get a Demo' to 'Free Demo' | +99.42% clicks | Greenlane Search Marketing |
| 7 | Plans and trial offer (saas, won 17 of 21) | 14-day paid-plan trial vs free plan | +268% paid sign-ups | Acuity Scheduling |
| 8 | How pricing is displayed (saas, won 16 of 19) | '10 GB' instead of 'unlimited' | +27.9% revenue | VWO |
| 9 | Navigation (saas, won 13 of 13) | 'Book a Demo' in mobile header | 8.30% to 9.88% demo visits | Copado |
| 10 | Testimonials and proof (saas, won 13 of 13) | Narrow demo form with logos and testimonials | +20% demo sign-ups (bundled) | POSist |
| 11 | CTA copy (lead gen, won 32 of 34) | 'Get a Free Quote' CTA | +60% leads | Bloom |
| 12 | Form fields (lead gen, won 23 of 26) | Drop-downs instead of free-text search | +57.25% leads | Supersonic |
| 13 | Headline (lead gen, won 19 of 20) | Customer quote as headline | +24.31% sign-ups | Laura Roeder Social Media |
| 14 | Page length (lead gen, won 14 of 14) | Shorter landing page | 1.02% to 1.77% conversion | Apexure |
| 15 | Multi-step form (lead gen, won 13 of 13) | Multi-step enquiry form | 0.96% to 8.1% conversion (before-after) | Venture Harbour |
| 16 | Paywall design (apps, won 26 of 28) | Two-step timeline paywall | +50% LTV | Pepper |
| 17 | Free trial length (apps, won 24 of 25) | 7-day trial instead of 14 | +8% install-to-paid | Zumba |
| 18 | Price display (apps, won 15 of 16) | Annual plan shown per month | +45% Brazil, +26% Mexico, +8% USA | Mojo |
| 19 | Plan mix (apps, won 14 of 18) | Weekly plan instead of monthly | +16% paying customers | Pixelcut |
| 20 | Paywall timing (apps, won 9 of 9) | Half as many, better-timed prompts | +81% on the primary offer flow; LTV up about 20% | Wizz |
The 20 A/B tests to run first
1. Ecommerce: Product title and headline copy (won 121 of 132)
The change: Rewrite the product title so it leads with the outcome, in an 'Achieve [outcome] with [product]' format, instead of the product name.
The result: +16.92% revenue per visitor, with conversion rate up 15.61%.
A/B test run by Dave Diederen. Sample size and the exact titles were not shared.
Why it likely worked: Shoppers scan the title first.
Naming the result they want answers 'is this for me?' before they reach the price.
Try it if: your product titles are just names or SKUs and your ads sell an outcome the page never repeats.
2. Ecommerce: Price and discount framing (won 104 of 126)
Blend CommerceThe change: Add live 'Save X% per oz' labels to the larger bag sizes on a coffee product page.
The result: +41% conversion rate for Stone Creek Coffee, with revenue per visitor up 26% and average order value down 10%.
A/B/C test run by Blend Commerce on one product page.
Why it likely worked: It shows the saving instead of asking shoppers to do the unit math.
The drop in order value is a reminder to watch more than one metric.
Try it if: you sell sizes, packs or quantities and the per-unit value of the bigger option is hidden.
3. Ecommerce: Navigation menu (won 102 of 128)
The change: Replace a plain list of category names in the mega menu with one line of description per category and a customer testimonial, with smaller service links.
The result: +$151,192/month (estimated) for a tennis and racket gear brand.
A/B test with three menus run by Oliver Kenyon.
A thumbnail version lost to it, and three things changed at once.
Why it likely worked: The menu is seen on almost every visit, so a small gain in clarity adds up.
Short descriptions help shoppers pick a category without guessing.
Try it if: you have many categories with names that mean little to a first-time visitor.
4. Ecommerce: Upsells and add-ons (won 101 of 112)
@stoykovallex1The change: Add a one-click 'Frequently Bought Together' bundle (tape and solvent) near add to cart on a golf grip page.
The result: +10.87% revenue per visitor for WedgeGuys, with conversion rate up 8.09%.
A/B test run by Alex Stoykov over 16 days on 12,421 visitors.
Why it likely worked: The add-ons are things the buyer needs to use the product, so the offer helps rather than distracts.
Try it if: your product needs accessories or refills that buyers often forget.
5. Ecommerce: Product image order (won 88 of 100)
@elevate_abThe change: Swap the first and second photos in a product gallery.
Nothing else changed and no new photos were shot.
The result: Revenue per visitor £3.32 to £3.60 for The Cornish Company.
A/B test run by Elevate A/B Testing over ten days on 9,913 visitors.
Why it likely worked: The first image does most of the selling on mobile, where many shoppers never swipe. Changing which photo leads costs nothing.
Try it if: you have a lifestyle or 'in use' photo buried in position three or four.
6. SaaS: CTA button copy (won 19 of 23)
The change: Change the homepage button from 'Get a Demo' to 'Free Demo'.
The result: +99.42% clicks for Corcentric, a B2B software company.
A/B test run by Greenlane Search Marketing in VWO.
About 800 visitors over two weeks, and the metric was clicks, not booked demos.
Why it likely worked: 'Free' removes the worry that a demo is a sales trap with a price tag.
One word is the cheapest test you can run.
Try it if: your main button asks for commitment without saying what it costs the visitor.
7. SaaS: Plans and trial offer (won 17 of 21)
The change: Send the main sign-up button to a 14-day full-feature trial of the paid plan, with card details up front, instead of a limited free-forever account.
The result: +268% paid sign-ups for Acuity Scheduling, who ran the A/B test in VWO. Free sign-up numbers were not reported.
Why it likely worked: Users who try the full product see the value of paid features before the trial ends.
This changed the offer, not just the page.
Try it if: a free plan gives away enough that few people ever upgrade.
8. SaaS: How pricing is displayed (won 16 of 19)
The change: Show '10 GB' of storage on the Business plan next to '2 GB' on Pro, instead of 'unlimited' on Business.
The result: +27.9% revenue for Powtoon. A/B test with three variants run by Wingify's services team, published by VWO. No sample size or duration was shared.
Why it likely worked: A concrete number is easier to compare than 'unlimited', which can read as vague or too good to be true.
Try it if: your pricing table uses fuzzy limits that buyers can't compare between plans.
9. SaaS: Navigation (won 13 of 13)
The change: Show a 'Book a Demo' button in the mobile header instead of hiding it inside the hamburger menu.
The result: 8.30% to 9.88% demo visits for Copado.
A/B test run by Copado in VWO. The source called it a 68.55% uplift, but those two rates work out to about 19%.
Why it likely worked: On mobile, anything inside the menu is out of sight.
Putting the main action in the header keeps it one tap away.
Try it if: your key conversion action is only reachable through the mobile menu.
10. SaaS: Testimonials and proof (won 13 of 13)
The change: Narrow the demo request form, explain what prospects receive, and add customer logos and testimonials beside it.
The result: +20% demo sign-ups (9.37% to 11.26%) for POSist.
A bundled change run by the POSist team in VWO, so the lift can't be credited to the testimonials alone.
Why it likely worked: Buyers about to hand over contact details want proof that companies like theirs got value.
Proof placed at the form works harder than proof on the homepage.
Try it if: your demo or contact page is a bare form with nothing around it.
11. Lead gen: CTA copy (won 32 of 34)
The change: Change the site-wide call to action from 'Book a Free Survey' to 'Get a Free Quote'.
The result: +60% leads for ADT UK's residential leads.
A/B test run by Bloom in VWO over 14 days on new visitors. The data dates from around 2013.
Why it likely worked: A quote is what visitors want; a survey sounds like a visit and a sales pitch.
The button now names the reward, not the chore.
Try it if: your CTA describes your process instead of what the visitor gets.
12. Lead gen: Form fields (won 23 of 26)
The change: Replace a free-text property search box with two drop-downs for apartment type and location.
The result: +57.25% leads for real estate agency Casa Mineira.
Split URL test run by Supersonic in VWO with about 7,500 visitors.
Why it likely worked: Choosing from a list is easier than typing, and it avoids searches that return nothing.
Try it if: a form field asks people to type something you could offer as choices.
13. Lead gen: Headline (won 19 of 20)
The change: Replace a marketing-style homepage headline with a reader's words: 'Yours is the only newsletter that I actually read.'
The result: +24.31% sign-ups for Laura Roeder Social Media.
A/B test run by the business itself in VWO over 12 days, with a reported 97% chance to beat the original.
Why it likely worked: A customer quote does the job of a value proposition and social proof at once, and it sounds less like an ad.
Try it if: you have a strong testimonial and a headline nobody would quote back to you.
14. Lead gen: Page length (won 14 of 14)
ApexureThe change: Replace a long landing page with a shorter one that sends paid visitors on to the same claim form.
The result: 1.02% to 1.77% conversion end to end for Reclaim My PPI Tax.
A/B test run by Apexure in Unbounce over about a month of paid traffic.
No traffic counts or confidence level were given.
Why it likely worked: For a simple, low-risk offer, extra copy delays the action without answering new questions.
Try it if: your offer is easy to understand and visitors arrive from ads already wanting it.
Complex or expensive offers may need the long page.
15. Lead gen: Multi-step form (won 13 of 13)
Venture HarbourThe change: Swap a single-page contact form for a multi-step enquiry form that asks easy questions first.
The result: 0.96% to 8.1% conversion for Venture Harbour's consulting enquiries.
This is a before-and-after result, not a split test, built over years, and Venture Harbour later sold a multi-step form tool.
Why it likely worked: Easy first steps get people started, and once they've started, they tend to finish.
Try it if: your form has more than five fields on one screen.
Test it properly as an A/B test rather than copying the number.
16. Apps: Paywall design (won 26 of 28)
The change: Turn a single-screen paywall into a two-step flow with a day-by-day timeline of the free trial.
The result: +50% LTV for Pepper, a recipe app, which ran the A/B test on 5% of users and then scaled it to over half.
Published as a Superwall case study, and Superwall sells the paywall tool.
Why it likely worked: A timeline answers the main trial fear: when will I be charged?
Less anxiety means more people start the trial.
Try it if: your paywall is a feature list with a price and a button.
17. Apps: Free trial length (won 24 of 25)
The change: Cut the free trial from 14 days to 7 days.
The result: +8% install-to-paid for the Zumba app, which ran the A/B test.
Published as a RevenueCat customer story; RevenueCat sells the experiments tool used.
Why it likely worked: A shorter trial gives users a nearer deadline, so more decide while they're still engaged.
Try it if: most of your trial users stop opening the app well before the trial ends.
18. Apps: Price display (won 15 of 16)
The change: Show the annual plan as a per-month price, billed annually.
The result: +45% Brazil, +26% Mexico, +8% USA in new revenue per paywall impression for Mojo, a video editing app, which ran the A/B test.
Why it likely worked: A monthly figure is easier to weigh against a monthly plan, and it makes the annual saving obvious. The lift varied a lot by country.
Try it if: you sell annual and monthly plans and show the annual total only. Check Apple's pricing display rules first.
19. Apps: Plan mix (won 14 of 18)
The change: Offer a weekly subscription instead of a monthly one on the paywall.
The result: +16% paying customers for Pixelcut, an AI photo editing app, which ran the A/B test.
Published as a RevenueCat customer story; prices weren't shared.
Why it likely worked: A small weekly price is an easier first yes for users who aren't sure how often they'll use the app.
Try it if: your app gets used in bursts and monthly plans feel like a commitment.
20. Apps: Paywall timing (won 9 of 9)
The change: Show about half as many paywall prompts, held back in low-intent moments and shown when users are more receptive.
The result: +81% on the primary offer flow; LTV up about 20% for Wizz, a dating app, which ran the A/B test.
The case study came from ContextSDK, which sells the timing tool.
Why it likely worked: Fewer prompts means less banner blindness, and each one lands at a moment when the user has just seen value.
Try it if: you show the paywall on every launch or after every action.
What the winners have in common
They change words before layouts
Titles, button text, headlines and price labels top every list.
Copy tests need no design work and can go live in a day. See more in the best copy tests.
They sit where everyone looks
The menu, the first product photo, the main CTA and the paywall are seen by nearly every visitor.
A test there reaches a sample size quickly, which matters when traffic is low. Our sample size guide explains why.
They make a cost or value concrete
'Free Demo', 'Save X% per oz', '10 GB' and a monthly price for the annual plan all replace something vague with a number or a promise.
Pricing tests win often for every type of business; see the best pricing page tests.
They remove effort
Drop-downs instead of typing, multi-step forms and shorter pages all ask less of the visitor. The best form tests has more.
What to test next: ranks 6 to 10
Once the first five are done, these are the next most frequent winners in each business type, with win counts. Browse our full idea list for more.
Ecommerce
- 6. Reviews: won 77 of 87
- 7. Filters and search: won 74 of 88
- 8. Shipping messaging: won 70 of 83
- 9. Hero section: won 69 of 76
- 10. CTA button: won 69 of 84
SaaS
- 6. Social proof: won 12 of 12
- 7. Page copy: won 11 of 17
- 8. Form fields: won 11 of 16
- 9. Page length: won 10 of 14
- 10. Headline: won 9 of 9
Lead gen
- 6. Page copy: won 12 of 15
- 7. Value proposition: won 12 of 15
- 8. Testimonials: won 11 of 13
- 9. Personalization: won 10 of 11
- 10. Pop-ups: won 8 of 8
Apps
- 6. Discounts: won 8 of 10
- 7. Freemium limits: won 6 of 8
- 8. Onboarding quizzes: won 5 of 5
- 9. Win-back offers: won 1 of 1
Our app tests only carry nine element tags, so this list stops at nine. The best paywall tests page goes deeper.
For more by type, see the SaaS A/B test results, lead gen A/B test results and best product page tests.
If you'd rather have someone plan and run these tests for you, Oliver Kenyon runs ecommerce tests like the mega menu winner above.
Real results
@oliverkenyon
@DaveDiederenWatch out for
- Bundled changes. The POSist and mega menu tests changed several things at once, so you can't tell which part won. Test one idea at a time when you can.
- Click metrics. The Corcentric and Copado results measured clicks and page visits, not booked demos. Track the step that makes money.
- Trade-offs. Stone Creek Coffee's conversion rose 41% while order value fell 10%. Check revenue per visitor, not just conversion rate.
- Vendor-published results. Many app and SaaS wins come from case studies by the testing vendor, which can lean positive.
- Before-and-after numbers. Venture Harbour's 0.96% to 8.1% is not a split test. Seasonality and traffic mix can create or hide a lift.
Not every winner repeats on another site. Our losing A/B tests page shows ideas that failed elsewhere, and A/B testing tools compares what to run them on.