The A/B tests from 2026 most worth learning from are the ones that broke a common rule and still had a clear number behind them.
At the top: Oliver Kenyon's cart drawer reassurance block (+23.8% revenue per visitor across 200,000+ visitors), Dave Diederen's tests showing that removing an email popup and adding countdown timers in European markets both lost, and BuyBye's 50% higher yearly price, which cut conversions but raised proceeds per user by 25%.
How we picked: we took every test in the LiftArchive database dated 2026 (222 when we wrote this, most of them ecommerce, plus mobile apps and a few lead gen sites).
We ranked them by how useful the lesson is to other businesses first and measured lift second.
Tests with a clear control, variant and published numbers ranked above vague ones.
We kept the losers, because a documented loss saves you a test.
We did not run these tests; each result is copied from whoever ran it.
A note on coverage: our archive has no SaaS tests dated 2026 yet, so this list covers ecommerce, apps and lead gen.
For SaaS, see our SaaS A/B test results.
For the full archive, start at LiftArchive research.
The 25 tests at a glance
| Rank | Test | Change | Result | Who ran it |
|---|---|---|---|---|
| 1 | Cart drawer reassurance block | Dispatch, returns and payment line under the checkout button | +23.8% revenue per visitor | Oliver Kenyon |
| 2 | Removed the email popup | Turned the popup off | -20.51% revenue per visitor (lost) | Dave Diederen |
| 3 | Countdown timer, European market | Added a countdown timer | -35.61% revenue per visitor (lost) | Dave Diederen |
| 4 | BuyBye yearly price +50% | Raised the yearly price | -19% conversions; +25% proceeds per user | BuyBye |
| 5 | $9.49 shipping fee | Tested shipping price levels | +43.81% revenue per visitor | Dave Diederen |
| 6 | Hero video to static image | Swapped homepage video for a benefit image | +52.67% revenue per visitor | Dave Diederen |
| 7 | "Buy 2 Get 1 Free" wording | Reframed a 2-pack at the same price | -52.38% revenue per visitor (lost) | Milos |
| 8 | $1.99 paid 30-day trial | Replaced monthly no-trial plan | D90 ARPU $2.23 → $2.89 (+30%) | Vahe Baghdasaryan |
| 9 | Hid reviews above the fold | Moved written reviews off the first screen | +23.05% revenue per visitor | Dave Diederen |
| 10 | Removed the bundle selector | One product, one price, one button | +7.8% revenue per visitor | Oliver Kenyon |
| 11 | Preu AI annual plan | Added an annual plan next to monthly | +80% proceeds per user; +22% trial conversion | Preu AI |
| 12 | Reviews on checkout | Showed customer reviews in checkout | +23.13% revenue per visitor | Dave Diederen |
| 13 | Skipped the cart | Sent shoppers straight to checkout | -6.58% revenue per visitor (lost) | Dave Diederen |
| 14 | BoldVoice hard paywall | Removed the paywall close button | More trial starts, but refunds rose by more than the gain (lost) | BoldVoice |
| 15 | Split size selector | Separated men's and women's sizes | +12.9% revenue per visitor | Oliver Kenyon |
| 16 | Insurance coverages page | Images, bullet previews, "Why us" section | +32% to quote pages | OuterBox |
| 17 | Best sellers above the fold | Products first, reviews lower on the homepage | +39.03% revenue per visitor | Dave Diederen |
| 18 | Pre-checked subscription | Subscription selected by default | -9.17% revenue per visitor (lost upfront) | Dave Diederen |
| 19 | PropGPT weekly price | $6.99 to $9.99 per week | +47.5% proceeds per user | PropGPT |
| 20 | Native page vs page builder | Rebuilt the same landing page in the theme | +20.9% revenue per visitor | Oliver Kenyon |
| 21 | Mercantil Bank phone-only form | Popup asking only for a phone number | +5.5% submissions, -23.6% CAC | Mercantil Bank |
| 22 | Stompers multi-page paywall | Feature screens before pricing | Single-page converted over 16% better (variant lost) | Stompers |
| 23 | No free-shipping threshold | Charged shipping on every order | +9.17% revenue per visitor | Dave Diederen |
| 24 | Free recipe e-book | Gift with knife purchases | +48.71% revenue per visitor | Oliver Kenyon |
| 25 | 14-day vs 7-day trial | Doubled the free trial length | Slightly more trial starts but fewer paid conversions (lost) | Daphne Tideman |
The top 25 A/B tests of 2026, ranked
1. A reassurance block under the cart drawer's checkout button (Oliver Kenyon)
The change: On a fashion store's cart drawer, Oliver Kenyon added a small block directly under the Checkout button: tracked next-day dispatch, 14-day easy returns, and the accepted card and Klarna logos.
A size dropdown and one other small fix shipped in the same variant.
The result: +23.8% revenue per visitor and +21.9% conversion rate at 99.4% confidence, over 16 days and 200,000+ visitors. A/B test.
Why it likely worked: It answers the last three questions people have (when will it arrive, can I return it, can I pay my way) at the moment they decide whether to click.
Try it if: You sell anything where fit or delivery time worries buyers, and your cart drawer currently shows only a total and a button.
Ideas for the rest of the drawer are in our cart drawer guide.
2. Removing the email popup (Dave Diederen)
The change: Dave Diederen turned off the email signup popup on a client's store, then repeated the test on a second store.
The result: It lost both times. On the recorded test: -20.51% revenue per visitor, -12.18% conversion rate and -9.49% AOV. A/B test.
Why it likely worked (for the popup): Dave Diederen's take: popups feel annoying to store owners, who see them every day, but the data doesn't show shoppers reacting the same way.
A popup also catches first-time visitors who would otherwise leave without buying.
Try it if: You want to remove a popup because it "feels annoying."
Test the removal instead of shipping it. See our popup tests for what else has worked.
3. A countdown timer in a European market (Dave Diederen)
The change: A site-wide countdown timer for a brand selling outside the US.
Dave Diederen says the team tested timers across 20+ brands in the two to three months before posting.
The result: Lost: -35.61% revenue per visitor, -17.35% conversion rate, -22.09% AOV. A/B test.
Why it likely worked (against the timer): Urgency doesn't read the same in every market.
The same team found a checkout timer won for US brands (+4.83% revenue per visitor), and Kasper Dolk saw removing a checkout timer win for a Dutch audience (+26.18% revenue per visitor) after a survey said it felt like too much pressure.
Try it if: You sell into Europe and run timers by default.
Split the test by market. More in our countdown timer tests.
4. A 50% higher yearly price (BuyBye)
The change: BuyBye, a spending-limit and savings app, raised its yearly price with a trial by 50% and kept the paywall design the same.
The result: -19% conversions; +25% proceeds per user.
A/B test, published as a Superwall case study.
Why it likely worked: The price rise outweighed the conversion drop.
Judged on conversion alone, it would have looked like a loss.
Try it if: Your app has never tested price.
Measure proceeds per user, not trial starts. Our subscription pricing guide covers how.
5. A $9.49 shipping fee (Dave Diederen)
The change: Dave Diederen tested several shipping price levels against the control; a $9.49 fee won.
The result: +43.81% revenue per visitor, +23.40% conversion rate, +16.54% AOV. A/B test.
Why it likely worked: The post doesn't give the control fee.
The lesson: shipping price is a lever you can test like any other price.
Try it if: You set your shipping fee once and never touched it.
6. Homepage hero video replaced with a static image (Dave Diederen)
The change: Three image variants (product focus, lifestyle, and a benefit-led shot) against a hero video.
Every image beat the video; the benefit callout did best.
The result: +52.67% revenue per visitor, +24.81% conversion rate, +22.32% AOV, on 38k visitors and 750+ orders. A/B test.
Why it likely worked: A benefit-led image says what you sell at a glance; a video makes visitors wait.
Try it if: Your homepage opens with a brand video.
7. "Buy 2 Get 1 Free" at the same price as a 2-pack (Milos)
The change: Milos (profile) compared "1-pack $155 / Buy 2 Get 1 Free $240" against "1-pack $155 / 2-pack $240." Same price, different framing.
The result: The "free" version lost: -52.38% revenue per visitor and -51.36% conversion rate. A/B test.
Why it likely worked (against the offer): Milos doesn't explain it.
One reading: three units is more than most buyers want.
Try it if: You're about to roll out a BOGO-style offer.
Test the framing before you commit. Our bundle tests have more.
8. A $1.99 paid 30-day trial (Vahe Baghdasaryan)
The change: Across 10 health and fitness apps, Vahe Baghdasaryan's agency, Tangent, replaced the monthly no-trial plan with a $1.99 paid trial for the first 30 days, next to the annual plan with a 7-day free trial.
The result: D90 ARPU $2.23 → $2.89 (+30%); conversion to paying 10.2% → 12.7%.
Pooled A/B test data; the post doesn't say how users were split.
Why it likely worked: A small upfront payment filters for serious users and gives a cheap way in.
Try it if: Your monthly plan converts poorly.
9. Hiding written reviews above the fold (Dave Diederen)
The change: Moved long written reviews out of the first screen of the product page.
The result: +23.05% revenue per visitor and +13.36% conversion rate.
Dave Diederen says the same change won 8 out of 10 times across brands. A/B test.
Why it likely worked: Long reviews push the price and button down.
A star rating near the title gives trust without the bulk.
Try it if: Your mobile product page shows review text before the buy button. See review tests.
10. Removing the bundle selector (Oliver Kenyon)
The change: For an 8-figure travel bag brand, Oliver Kenyon replaced a 1/2/3-bag bundle selector with a tighter layout: one bag, one price, one button.
The result: +7.8% revenue per visitor, +8.7% conversion rate, +6.7% profit per visitor, with AOV unchanged. A/B test.
Why it likely worked: Most people buy one bag, so a quantity choice is a needless decision.
Try it if: Your product is a one-per-customer purchase and you added bundles because they work for consumables.
11. An annual plan priced against monthly (Preu AI)
The change: Preu AI, an exam prep app in Chile, went from a monthly-only paywall to monthly plus an annual plan priced at 60% of the monthly price.
The result: +80% proceeds per user; +22% trial conversion. A/B test (Superwall case study).
Why it likely worked: The monthly price makes the annual plan look like a deal.
Try it if: Your paywall offers a single plan.
12. Customer reviews on the checkout page (Dave Diederen)
The change: A few strong customer reviews shown in checkout.
The result: +23.13% revenue per visitor, +17.96% conversion rate, +4.39% AOV. A/B test.
Why it likely worked: Checkout is where last-minute doubt stops people, and reviews answer it there.
Try it if: Your platform lets you add content to checkout. More in our best checkout tests.
13. Skipping the cart (Dave Diederen)
The change: Sent shoppers straight from add-to-cart to checkout.
The result: Lost: -6.58% revenue per visitor and -10.23% conversion rate, though AOV rose +4.08%. A/B test.
Why it likely worked (for the cart): Dave Diederen calls the cart "your last sales page," where upsells and reassurance live.
Try it if: Someone proposes "fewer steps" as an automatic win. Test it first.
14. A hard paywall with no close button (BoldVoice)
The change: BoldVoice, an accent coaching app, removed the X button from its onboarding paywall.
The result: More trial starts, but refunds rose by more than the gain.
No figures given. A/B test, from a RevenueCat podcast write-up.
Why it likely worked (against the change): Forced trials bring in people who didn't want to pay, and they ask for refunds.
BoldVoice now judges tests on net revenue per exposed user on matured cohorts.
Try it if: You're weighing a hard paywall.
Track refunds, not just trials. See hard vs soft paywalls.
15. Separate men's and women's sizes (Oliver Kenyon)
The change: For an apparel brand, Oliver Kenyon redesigned the size selector so men's and women's sizes were separate. No new offer, no discount.
The result: +12.9% revenue per visitor, +10.4% conversion rate, +12.8% profit per visitor, +$38,699/month projected. A/B test.
Why it likely worked: A mixed size list makes people work out which size is theirs, and some give up or pick wrong.
Try it if: Your session recordings show people tapping sizes back and forth.
16. An insurance coverages page with real previews (OuterBox)
The change: OuterBox (profile) replaced small icons with real images and bullet previews for each insurance product, simplified the heading and added a "Why us" section.
The result: +32% to quote pages (+32.14% progression).
50/50 split over about four weeks. A/B test, published as a Convert case study.
Three changes were tested together, and the metric is progression, not completed quotes.
Why it likely worked: Visitors could see what each policy covered before clicking.
Try it if: You run a lead gen site with a services page made of icons and one-word labels. See lead gen test results.
17. Best sellers above the fold on the homepage (Dave Diederen)
The change: Moved the best-sellers row to the first screen and pushed the reviews section lower.
The result: +39.03% revenue per visitor, +21.75% conversion rate, +14.20% AOV. A/B test.
Why it likely worked: Homepage visitors are already shopping and want products, not quotes.
Try it if: Your homepage opens with testimonials or brand story.
18. Subscriptions with the option pre-checked (Dave Diederen)
The change: Introduced a subscription option on the product page and selected it by default.
The result: Lost upfront: -9.17% revenue per visitor, -2.86% conversion rate.
But subscription share of orders rose +1,295%. A/B test.
Why it matters: Dave Diederen's point: a subscription test can look worse upfront and still be the better business decision.
Try it if: You sell consumables. Decide your success metric (first-order revenue or lifetime value) before the test starts.
19. A higher weekly price (PropGPT)
The change: PropGPT, a sports betting insights app, raised its weekly price from $6.99 to $9.99 with the same images and copy.
The result: +47.5% proceeds per user. The conversion drop was described as minimal but not quantified. A/B test (Superwall case study).
Why it likely worked: Buyers of niche, high-intent apps are often less price-sensitive than founders expect.
Try it if: You priced your app by copying competitors. See our best paywall tests.
20. A native landing page vs the same page in a builder (Oliver Kenyon)
The change: Oliver Kenyon's team rebuilt a client's best landing page natively in Shopify with Liquid and compared it with the identical page built in a popular landing page builder. Same copy, images, structure and offer.
The result: +20.9% revenue per visitor, +9.9% conversion rate, +10.0% AOV. A/B test.
Why it likely worked: Only the build changed, so speed and cart integration are the likely reasons.
Try it if: Your paid traffic lands on builder pages.
21. A phone-only popup form (Mercantil Bank)
The change: Mercantil Bank, a Brazilian bank offering loans to people 50+, replaced a multi-field form with a popup asking only for a phone number, plus copy explaining what the loan simulation would show.
The result: +5.5% submissions, -23.6% CAC, and +15.8% qualified lead rate.
A/B test with a phased rollout, published by the testing vendor Croct; no sample size given.
Why it likely worked: One field is easy, and the copy told people what they'd get in return.
Try it if: Your lead form asks for details your sales team collects on the first call anyway.
22. A multi-page paywall vs a single page (Stompers)
The change: Stompers, a gamified walking app, tested a multi-page flow (feature screens and a trial reminder before pricing) against its single-page paywall.
The result: The multi-page variant lost. The single page converted over 16% better, with more yearly trials.
Price levels also varied. A/B test (Superwall case study).
Why it likely worked (for the single page): Extra screens give people more chances to leave before they see the offer.
Try it if: You're copying a long multi-step paywall from another app. Multi-step flows win elsewhere, so test it.
23. Dropping the free-shipping threshold (Dave Diederen)
The change: Removed the free-shipping threshold and charged shipping on every order.
The result: +9.17% revenue per visitor, +7.02% conversion rate, +2.01% AOV. A/B test.
Why it likely worked: The post doesn't say.
One possibility is that a threshold set too high made the fee look like a penalty, while a flat fee is easy to accept.
Try it if: Your threshold was set by guesswork.
Pair it with test #5. See free-shipping threshold tests.
24. A free recipe e-book with knife purchases (Oliver Kenyon)
The change: For a client selling cooking knives, Oliver Kenyon added a free BBQ recipe e-book with each purchase.
The result: +48.71% revenue per visitor. A/B test.
Why it likely worked: A digital gift costs almost nothing and ties the product to what buyers want to do with it.
Try it if: You're about to offer a discount. Try a relevant free extra first.
25. A 14-day trial instead of 7 days (Daphne Tideman)
The change: Daphne Tideman (profile) tested a 14-day free trial on a wellness app with long-form content, to match its 14-day activation window.
The result: Lost: slightly more trial starts but fewer paid conversions.
The app went back to 7 days. No numbers given. A/B test.
Why it likely worked (against the change): A longer trial gives people more time to forget about the app before the charge.
Try it if: You think a longer trial will fix weak activation. Fix onboarding first.
What the winners have in common
Most "best practices" depend on context
Popups, countdown timers, bundles, free shipping and reviews above the fold all produced results that cut against the usual advice.
Timers won for US brands and lost in Europe.
The lesson from 2026 is the same as every year: test the rule on your own traffic.
Our best practices that failed list collects more.
Reassurance at the decision point beats more persuasion earlier
The cart drawer block (#1), reviews in checkout (#12) and the clear size selector (#15) all answer a doubt at the moment someone is about to commit.
Price is underused as a test
Four entries (#4, #5, #19, #23) are straight price or shipping fee tests, and all four raised revenue per visitor or per user.
The metric decides the winner
BuyBye lost on conversions and won on proceeds.
BoldVoice won on trial starts and lost after refunds.
The subscription test lost on first-order revenue and won on subscription share.
Pick revenue per visitor, proceeds per user or net revenue before you start.
What to test next
- Your shipping fee and free-shipping threshold, as their own price test.
- A reassurance line (delivery, returns, payment) under your cart or checkout button.
- Your hero: one benefit-led static image against whatever is there now.
- Any urgency element, split by market if you sell internationally.
- For apps: one price increase, judged on proceeds per user after refunds.
- For lead gen: cut your form to the one field you actually need first.
If you're new to testing, start with our first A/B tests to run.
More 2026 results from the archive
@oliverkenyon@oliverkenyon
@DaveDiederen
@CROguy_
@kasper_dolkWatch out for
Single-test wins: Most results come from one test on one site, reported by whoever ran it.
Treat them as ideas to test, not changes to ship.
Bundled changes: #1, #16 and #21 changed more than one thing at once.
Fake urgency: Timers that reset or false "selling fast" banners can win a short test and cost trust later.
Short windows: Price, subscription and trial tests need time for refunds and renewals to show up.
Judge them on matured cohorts, as BoldVoice now does.
If you want help choosing and running tests like these, Oliver Kenyon ran five of the 25 on this list.