SaaS A/B Test Results: What SaaS Companies Test and What Wins

Across the 205 SaaS and subscription tests in LiftArchive, 172 (84%) are reported as wins, 23 as losses and 10 as flat.

The most-tested areas are CTAs, sign-up forms, copy and pricing pages, and the cleanest wins come from removing friction from sign-up and showing proof near the form.

Treat the win rate as a ceiling: companies publish wins far more often than losses.

What SaaS companies test, by the numbers

We pulled every result tagged saas from the LiftArchive test database in November 2026. That gave n = 205 tests credited to 93 different companies and agencies.

The tag covers software, B2B services and subscription businesses such as news sites and online courses, so it is broader than pure SaaS.

Each test has a page type, element tags, a direction (won, lost or flat) and an evidence type.

"A/B test" means one change against a control.

"Bundled" means several changes shipped together, such as a full landing page rebuild.

"Before-after" means the change was compared with an earlier period, not a live control.

The pattern to notice: the weaker the evidence, the higher the win share.

Single-change A/B tests win 77% of the time in our data, while bundled redesigns and before-after comparisons almost never lose.

That is not because redesigns are safer. It is because agencies publish their redesign wins as case studies and rarely publish the misses.

By page type, landing pages dominate. A large share of those 51 are agency rebuilds of paid-search landing pages, 24 of them from KlientBoost alone.

Checkout is where SaaS tests lose most often: 4 of 13. Payment flows have little slack, and two of those losses came from changing how card payment was offered.

Compare this with our lead-gen results, where almost nothing is reported as a loss.

Which themes win, and which ones lose

To group tests by theme, we mapped each test's element tags to ten themes.

A test can count in more than one theme (a pricing page redesign with testimonials counts under both pricing and social proof), so the rows add up to more than 205.

Copy and headline tests are the most honest coin flip in the set.

Only 9 of 17 single-change copy tests won.

Pricing is similar: 16 wins and 5 losses out of 22. These are the areas where SaaS teams have strong opinions and the data often disagrees.

Social proof, navigation and onboarding show no published losses at all.

Read that as "people publish these wins," not "these never fail."

For a wider view across verticals, see the most-tested elements.

Sign-up friction: the most repeatable wins

The most consistent SaaS pattern is removing a reason not to sign up.

Teamleader ran a series of single-change tests on its sign-up page: a no-credit-card note added 12.5% more sign-ups, and customer logos moved conversion from 36% to 40.3%.

Kareo Marketing cut sign-up fields and got 30% more new doctor sign-ups.

Mida added testimonials to its sign-up page for a 56.94% lift in sign-up completion.

But more steps or cleverer forms can backfire.

MarketingExperiments reports that a three-step free trial flow for RegOnline cut completed leads by 24.5%.

Patrick McKenzie's Mad Libs-style sign-up form lowered trial sign-ups by 22%.

Our SaaS sign-up page collection has the full set.

Pricing pages: big wins, frequent losses

Pricing tests produce some of the largest reported numbers.

Server Density's move to packaged pricing lifted revenue 114%.

Meebox's two-year plan discounts raised revenue 121.56%. VWO's case study on Powtoon's pricing page storage values reports +27.9% revenue.

Losses are just as instructive. In a MarketingExperiments test, removing the term-length dropdown from a subscription checkout cut conversion by about 40%.

Conversionista's first Fortum test, price blocks in a modal, lost 9% click-through; the dropdown iteration then won 9%. See SaaS pricing page tests for more.

Trials and onboarding: measure past the sign-up

Onboarding tests report activation, not clicks, which makes them more useful.

AURUM's onboarding checklist raised activation 187%. Patrick McKenzie cut the free trial limit from 15 to 8 and paid conversion went from 1.04% to 1.89%.

The warning comes from the same source. Patrick McKenzie's 2008 redesign brought more trials, but sales roughly halved.

The 2012 redesign added 10% more trial sign-ups while purchases stayed flat.

A SaaS test that stops at sign-ups can call a revenue loss a win.

Our SaaS onboarding tests and free trial tests track these deeper metrics where they were reported.

A

→

B

+12.5% sign-ups

'No credit card required' on trial sign-up

Signup pageTeamleader

A

→

B

+114% revenue

Packaged pricing tiers instead of per-server

Pricing pageServer Density

A

→

B

+187% activation

Floating onboarding checklist in a trial

OnboardingAURUM

A

→

B

+30% new doctor sign-ups

Fewer fields on doctor sign-up form

Signup pageKareo Marketing

A

→

B

more trials, but sales roughly halved

Full site redesign and merger

Site-widePatrick McKenzie

A

→

B

-24.5% completed leads

Three-step free trial sign-up on the homepage

HomepageMarketingExperiments

How big are SaaS wins?

We took every winning SaaS test whose headline result is a plain percentage lift (for example "+30% sign-ups") and calculated the median.

Results stated as before/after rates, dollar amounts or words were left out.

Do not plan a roadmap around a 34% median.

These are published winners only, many tracked clicks rather than revenue, and small tests that happen to win tend to overstate the effect (the winner's curse).

Read the median as "what gets published," not "what you should expect."

Our average lift research covers this across all verticals.

Some lifts in the set are flagged by the people who ran them.

In one MarketingExperiments test, a +445% lift was not trusted and is recorded as flat.

Why public SaaS results skew toward wins

The 84% win share in our SaaS data is far above what experimentation teams report from their own programs.

In a Microsoft paper, Ronny Kohavi and co-authors wrote that "only about 1/3 of ideas improve the metrics they were designed to improve" (Kohavi et al., Online Experimentation at Microsoft).

The gap is publication bias. Agencies publish case studies to win clients, and vendors publish customer stories to sell tools.

One source in our SaaS set is close to a full log: Patrick McKenzie's Bingo Card Creator tests.

Of those 23 tests, 11 won, 7 lost and 5 were flat, a 48% win share.

Two more counting notes. Across the full database (all verticals), single-change A/B tests win 79.5% of the time in ecommerce, 77% in SaaS, 82% in apps and 93% in lead gen.

And our win share counts the result the source reported, not one we re-checked; many sources did not publish sample sizes.

See A/B test win rates and losing tests for the wider picture.

What to take from this for your own SaaS tests

• Start with sign-up friction. Fewer fields, a no-credit-card note and proof next to the form are the most repeated wins in the set.
• Expect pricing and copy tests to lose often. About a quarter of single-change pricing tests and nearly half of copy tests lost or were flat.
• Measure paid conversion, not just trials. The Bingo Card Creator redesigns show trial gains that did not turn into sales.
• Discount bundled redesign wins. They lose rarely in public data because the losses are not published, and they don't tell you which change worked.
• Plan for a lower win rate than case studies suggest. A program that wins a third to half of its tests is in line with the most complete logs we have.

For conversion baselines to test against, see SaaS conversion rate benchmarks.

For all our original studies, go to LiftArchive research.