Every A/B test starts with the same practical question: how do you actually split the traffic? The standard answer is a JavaScript snippet — a testing tool that loads on your page, decides which version the visitor should see, and swaps the content around. That works, but it only works on your website, and it carries baggage: an extra script, a flash of the original page, and a setup step for every single experiment.
There’s a simpler layer to run the split: the link itself. A split link is one short URL that distributes clicks across two or more destination pages in whatever ratio you choose. The visitor just clicks; the redirect decides. This guide covers how link-level A/B testing works, where it beats script-based tools, and how to set up your first split in about a minute.
What is link-level A/B testing?
A normal short link points at exactly one destination. A split link holds a small list of variants — each one a destination URL with a weight and an optional label — and assigns every click to one of them, in proportion to the weights. Two variants weighted 1 and 1 give you a clean 50-50 test. Weights of 9 and 1 send roughly 90% of clicks to the first page and 10% to the challenger — the right shape for trying a risky redesign without betting the whole campaign on it.
On 302.sh a split link can hold between 2 and 10 variants, so you’re not limited to a two-way test: an A/B/C test of three landing pages is the same one-minute setup.
Crucially, the assignment happens inside the redirect, at the edge, before your page begins to load. The visitor never sees a decision being made — they click a link and land on a page, exactly like any other click.
Where script-based testing tools stop
Client-side testing tools assume the experiment lives on a page you control and can add JavaScript to. That assumption quietly excludes most of the places marketing actually happens:
- Email and SMS. You can’t run a script inside an inbox. A split link A/B tests the landing page for a campaign straight from the message.
- QR codes and print. A QR code on packaging or a poster resolves to one URL, forever. If that URL is a split link, the printed code becomes a running experiment.
- Social bios and marketplaces. One link slot, no code access. A split link turns it into a test.
- Paid ads. Test two landing pages against one ad creative without touching the ad platform’s own (often expensive) experiment features.
There’s also a performance argument. Script-based tools have to load, evaluate, and often repaint the page — the infamous “flicker” where the original version flashes before the variant appears. A redirect-level split adds none of that: each variant is just a normal page load of a normal URL.
Split links vs. testing scripts at a glance
| Split link | JS testing tool | |
|---|---|---|
| Setup | Add URLs + weights to a link | Install snippet, build experiment |
| Works in email / QR / print | Yes — anywhere a link works | No — needs your page’s JS |
| Page-load flicker | None — normal page load | Common — content swaps in |
| What you can test | Whole pages / destinations | Elements within one page |
| Best for | Campaigns, landing pages, offers | Fine-grained on-site UX tweaks |
The two approaches are complements, not rivals: a testing script is the right tool for changing a button inside a page you own, while a split link is the right tool for testing which page the click should land on — especially from channels where no script can follow.
Sticky variants: one visitor, one experience
A naive random split has an ugly failure mode: the same person clicks twice, sees two different pages, and either gets confused or pollutes your data by converting on the “wrong” variant. 302.sh avoids this with a variant cookie: the first click assigns a variant, and every later click from the same visitor lands on the same page.
Consistency is what makes the numbers trustworthy. If assignment were re-rolled on every click, repeat visitors would smear across variants and neither side of the test would mean anything.
Measuring the winner
The link’s analytics record which variant every click was assigned to, so the traffic side of the test — how many clicks each page received — is visible right in the dashboard. Conversion, the thing you actually care about, is measured where it happens: on the destination pages. Two practical patterns:
- Distinct URLs. If the variants are genuinely different pages (
/pricing-avs/pricing-b), your existing web analytics already separates them. - UTM parameters. If both variants point at the same page with different content served server-side, append different UTM parameters (e.g.
?utm_content=variant-a) so conversions attribute cleanly.
And read the results honestly. Give the test enough clicks that the difference can’t be noise — a few dozen clicks per variant proves nothing. Change one meaningful thing per test, let it run its course, then ship the winner. (If you’re benchmarking what click-through you should expect in the first place, see What Is a Good CTR? Click-Through Rate Benchmarks by Channel.)
Setting up a split link on 302.sh
Creating a test takes about a minute:
- 1. Create a new link and set the default destination as usual.
- 2. Open the Smart routing section and choose “A/B split”. Add a row per variant: weight, destination URL, and an optional label (up to 64 characters) so the analytics read “spring-hero” instead of “variant 2”.
- 3. Set the weights. Equal weights for a fair test; lopsided weights (like 9/1) to trial something risky on a small slice of traffic.
- 4. Share the one short link everywhere the campaign runs. Every channel feeds the same experiment.
Because 302.sh serves every short link as a 302 (temporary) redirect, you can edit weights, swap variant URLs, or end the test and point 100% of traffic at the winner — and the change is live instantly for every visitor. Nothing is cached against you.
A/B split routing is available on the Pro ($29/mo) and Business ($79/mo) plans. Redirects are unlimited on every plan, and every plan — including free — comes with 90-day click analytics.
Frequently asked questions
Do I need to install any code to A/B test with a short link?
No. The split happens inside the redirect itself, at the edge, before your page even starts loading. There is no JavaScript snippet to install, no tag manager to configure, and no flicker while a testing script swaps content. If a channel can carry a link — email, SMS, a QR code, a bio, a paid ad — it can carry the test.
How are visitors split between variants?
Each variant gets a positive integer weight, and clicks are distributed in proportion to those weights. Two variants weighted 1 and 1 split traffic 50-50; weights of 9 and 1 send roughly 90% of clicks to the first URL and 10% to the second. A split link can hold between 2 and 10 variants.
Will the same person see a different page on their next click?
No. A variant cookie remembers which variant a visitor was assigned, so a returning visitor lands on the same page every time. That keeps the experience consistent and your data cleaner — repeat clicks don’t leak between variants.
How do I measure which variant converts better?
The link’s click analytics record which variant every click was assigned to, so you can compare traffic volumes per variant. Conversion itself is measured on your destination pages: point each variant at a distinct URL or append different UTM parameters, and your web analytics will attribute signups or purchases to the right variant.