The fastest way to grow most marketing isn’t a big rebrand — it’s a steady habit of testing one change at a time and keeping what wins. That’s all a/b testing is: showing two versions to comparable audiences, measuring which performs better, and letting data settle arguments that opinion can’t. Done consistently, it compounds. This guide covers how to run A/B tests that actually tell you something, with a Singapore lens.
What a/b testing really is
A/B testing (also called split testing) compares two versions of something — version A, the control, and version B, the variant — by splitting your audience and measuring which drives more of the outcome you care about. That outcome might be clicks, sign-ups, purchases, or enquiries.
The discipline of a/b testing matters because marketing is full of confident guesses. Does a red button beat a blue one? Does “Get a quote” beat “Contact us”? You can argue about it, or you can test it and know. The point isn’t any single test — it’s building a habit of small, evidence-based improvements.
What’s worth testing
Test things that sit close to the decision and have enough traffic to produce a clear result. High-impact candidates include:
- Headlines and value propositions — usually the biggest lever on a page.
- Calls to action — wording, colour, placement of the button.
- Ad creative — image vs video, different hooks, different first frames.
- Email subject lines — the gate to everything else in the email.
- Landing page layout — form length, social proof placement, hero image.
- Offers — free delivery vs a discount, in SGD terms your audience responds to.
Don’t test trivial things on low traffic. Testing a button shade on a page with a handful of weekly visitors will never reach a reliable answer.
Test one variable at a time
The cardinal rule: change one thing per test. If you change the headline and the image and the button, and B wins, you’ve no idea which change did it — and you can’t repeat the success.
Isolating a single variable keeps results clean and learnings reusable. If you genuinely need to test many combinations at once, that’s multivariate testing, and it requires far more traffic to be valid. For most Singapore SMEs, simple one-variable A/B tests are the right tool.
How to run a valid test
A test is only useful if you can trust the result. Follow a basic structure:
- Form a clear hypothesis. “Changing the CTA from ‘Submit’ to ‘Get my free quote’ will increase form completions, because it’s clearer and lower-risk.”
- Pick one primary metric. Decide upfront what success means, so you’re not fishing for any number that looks good after the fact.
- Split traffic evenly and randomly. Both versions should run at the same time to the same kind of audience.
- Run long enough. Let the test gather a meaningful sample and run across full weeks — weekday and weekend behaviour in Singapore differs.
- Wait for significance. Don’t call a winner the moment one version edges ahead.
Avoid the false wins
Most botched tests fail in the same predictable ways. Watch for these:
- Stopping too early. Early leads often vanish as more data comes in. Decide your sample size or duration before you start, and stick to it.
- Too little traffic. Small samples produce big swings that mean nothing. If volume is low, test bigger, bolder changes that can show a clear difference.
- Testing during anomalies. A test running over a major sale period, a public holiday, or a campaign spike won’t reflect normal behaviour.
- Chasing tiny differences. A fraction-of-a-percent gap inside the noise isn’t a real win.
- Ignoring losers. A test that “fails” still teaches you what your audience doesn’t want — that’s valuable.
Build a testing habit
The brands that win aren’t running one heroic test — they’re running a steady stream of them and stacking the gains. Keep a simple log: hypothesis, what you tested, the result, and what you learned. Over months, small validated improvements to ad creative, landing pages and emails compound into a meaningfully better-performing funnel.
This is core to how we run performance marketing — continuous testing of ads, audiences and landing pages so budget flows to what proves itself. It also informs the creative work: the winning angles from ad tests shape the next round of video content, so you’re producing more of what already converts rather than guessing.
Frequently asked questions
How long should an A/B test run?
Long enough to gather a meaningful sample and to cover at least one or two full weeks, so weekday and weekend behaviour are both captured. The exact duration depends on your traffic — higher-volume pages reach a reliable result faster. The key discipline is deciding the duration or sample size before you start, not stopping the moment a version looks ahead.
How much traffic do I need for a/b testing?
Enough that the result isn’t just noise. There’s no universal number — it depends on how big a difference you’re trying to detect and your current conversion rate. If your traffic is modest, test bolder changes (a whole new headline or offer) rather than subtle tweaks, since larger effects show up with smaller samples.
What if my A/B test shows no clear winner?
That’s a valid and common outcome. It usually means the change wasn’t significant enough to matter, which is itself useful — keep the simpler or cheaper version and move on to a bolder test. Not every test produces a winner, but a steady cadence of tests reliably produces gains over time.
Want a testing programme that keeps improving your results? Tell us about your goals and we’ll build a plan around what your data proves.
