LiftArchive collects real A/B test results that practitioners and companies have already shared in public, and credits every one to whoever ran it.
Sources
Tests come from three kinds of public sources: posts by practitioners on X, case studies on agency and testing-tool websites, and company blogs.
Every test page names its source and links to the original.
We follow each site's robots.txt and terms, skip anything behind a login, paywall or captcha, and leave out sources whose terms forbid reuse.
What counts as a result
A/B test: a split test with a measured result, the gold standard.
Multivariate test: several versions tested at once.
Bundled redesign: many changes tested together, so the result can't be pinned on one element; these are labeled.
Before and after: a change measured over time rather than split; also labeled, and left out of rankings like the biggest wins.
Idea: a suggestion with no result yet; shown separately and never counted in win rates.
A test shared as not significant counts as flat, not as a win.
How we write summaries
Every summary, mockup and analysis is written by us, in our own words.
We never copy sentences or rehost images; original posts are embedded or linked.
Numbers are reported exactly as the source states them, and a test with no number is marked as unquantified.
Credit and permission
Each test page shows who ran the test, with a link to their profile and the original post, and our commentary is labeled as ours.
Takedowns
If you shared a test that appears here and want it corrected or removed, email [email protected].
We act within 24 hours.
What gets indexed
Test pages are indexed once they have a full write-up.
Tests with no reported number stay out of search results, and hub pages need at least three tests.