Most A/B testing advice on the internet is written for people running enterprise marketing teams with 6-figure ad spend. None of it is practical when you're a creator with 15,000 followers running a Squarespace shop in your spare time.
Linkstacked's A/B link testing is the version for the rest of us. You create two variants of a link, we serve them at random to your visitors, we measure clicks (and optionally downstream conversions), and once the difference is statistically significant we tell you which one to keep. The whole thing is a 90-second setup and runs in the background while you do your actual job.
What you can test
Practical things, in roughly the order you should test them:
- 1Button copy. 'Buy now' vs. 'Get yours' vs. 'Pre-order today'. Smallest effort, biggest CTR swings — copy changes routinely move CTR by 15-40% in our data.
- 2Headline / preview text under the link. Same destination, different framing. Often outperforms button-copy changes for niche audiences.
- 3Button colour. Hero-link colour swap (your accent vs. neutral). Smaller effect than copy but easy to run alongside copy tests.
- 4Destination URL. Two different sales pages — same buyer intent, different layouts — and Linkstacked routes 50/50 to compare conversion.
- 5Position. Sometimes the test isn't 'which link', it's 'which slot'. Run a duplicate of the same link in two different positions and compare effective CTR.
Tip
Don't test more than one variable at a time. If your two variants differ in both copy and colour, you won't know which one moved the needle. The whole point of A/B is isolating cause.
How we decide a winner
Most low-end A/B tooling 'picks a winner' as soon as one variant has more clicks. That's statistical nonsense — small samples bounce around for noise reasons that have nothing to do with the variants. You can declare a winner that's actually no better than its opponent.
We use a sequential significance test (specifically, a Bayesian variant) that accounts for the running sample size. The test running tells you, in plain English: 'Variant B is currently 12% better than Variant A, with 87% confidence. We'll keep running until we hit 95% confidence or 14 days.' That second number is your bound — we never run forever, and we never declare a winner you can't trust.
Minimum traffic for a meaningful test
Honest answer: you need at least about 200 clicks per variant to detect a real 20% improvement. That's the actual minimum. Below that, the test won't converge — we'll tell you 'not enough data' rather than fabricate a winner. If your page gets 50 clicks a day total, an A/B test takes about 2 weeks to converge. That's not a bug, that's statistics.
If your page is doing under 30 clicks/day, A/B testing isn't your highest-leverage investment yet — work on traffic and content first, then come back to A/B testing once you have the volume to test cleanly.
Conversion tracking, not just clicks
On the Build plan we extend A/B testing to track downstream conversions, not just clicks. So if your variants both link to the same Stripe checkout but with different button copy, we can compare 'clicked AND purchased' rather than just 'clicked'. This matters because click-bait variants often win on CTR but lose on conversion — you can't see that with click-only testing.
Conversion tracking integrates with Stripe, Shopify, Patreon, Gumroad, Cal.com, and (with a tiny snippet) any custom destination. Setup is documented in detail but most creators don't even need to read it — for the supported integrations it's a single toggle in the link editor.
“I would have sworn my black 'Get the preset' button was better than my orange 'Try it free' one. The data says I was wrong by 28%. Six weeks later, orange button + free trial copy is the new default across my whole shop. The reason for that was a single A/B test that I almost didn't bother running.”
Things to watch out for
- Don't peek and call it early. The whole point of significance testing is to wait for the threshold. Calling it on day 3 because B is currently winning is statistical malpractice — it'll bite you with bad decisions over time.
- Don't test what doesn't matter. If your two variants are 'Buy now' vs. 'Buy now!' the lift is probably 1-2% and you'll need 50,000 clicks to detect it. Save your tests for changes you actually expect to move the needle.
- Don't run more than 2 simultaneous tests on the same link. The two tests interact and your results get muddy.
- Mind seasonality. A test that runs through Black Friday will measure 'works on a sale weekend' more than 'works in general'. If you can avoid it, time tests for normal traffic windows.
Plan differences
A/B testing is a Build-plan feature ($39/mo). The reason it's not on lower plans is that the statistical infrastructure (live confidence intervals, multi-day evaluation, conversion attribution) is genuinely expensive to operate. If A/B testing is the only thing keeping you off Build, the math usually pays for itself with one or two well-chosen tests.
Run your first test today
If you're already on Build, your first test should be on your hero link's button copy. Take your current copy, brainstorm one alternative that frames the offer differently ('Get yours' vs. 'Try free for 30 days' vs. 'Read the first chapter' — whatever fits your offer). Set up the test, walk away for a week. The data will tell you which version of yourself was right.
Share this with a teammate evaluating Linkstacked.