On this page
What structured testing means
An A/B test compares two or more versions of something, such as an ad, an audience or a landing page, to see which performs better against a defined goal. The idea is simple. The discipline is harder: most “tests” in ad accounts change several things at once, stop after a couple of days, or are decided by whoever likes an ad more.
We run tests with a clear hypothesis, one variable at a time, and enough budget and duration to reach a conclusion. Losing variants are turned off on evidence rather than instinct, and every result is recorded so the next campaign starts from what was learned.
What we test
- Creative: hooks, angles, formats, visuals and video openings.
- Copy: headlines, offers, calls to action and length.
- Audiences: broad versus defined, different lookalike seeds, interest groups or job functions.
- Landing pages: headlines, form length, layout, proof elements and page speed changes.
- Lead capture method: instant form versus landing page versus WhatsApp chat.
- Bidding and structure: where data volume allows a fair comparison.
Testing the biggest levers first
Not all tests are equally valuable. Changing the offer or the core message can change results substantially, while tweaking a button label rarely does. We start with bold tests, such as a different offer, a different angle or a different lead capture method, and move to finer details once the big questions are answered. This matters most for smaller budgets, where only large differences can be measured reliably.
Who benefits
Testing pays off for any business that plans to advertise for more than a few weeks. A D2C brand learns which product angle sells. A coaching institute learns whether parents or students respond better to a message. A real estate agency learns whether site-visit offers or brochure downloads bring better leads. A clinic learns whether a treatment explainer or an appointment offer works better. Over months, those lessons compound.
Testing also protects you from confident opinions. Everyone in a business has views on which ad or headline is best, and structured tests let the audience decide instead.
How we run tests
- Prioritise. Start with tests most likely to matter, usually offer and creative angle, before small details.
- Write a hypothesis. For example: a problem-first hook will lower cost per qualified lead compared with an offer-first hook.
- Isolate one variable. Everything else stays the same.
- Use the right tool. Platform experiment features where available, or carefully matched campaigns and ad sets.
- Run to a conclusion. Enough conversions and time to smooth out day-of-week and random swings.
- Decide and record. Winners scale, losers stop, and the result goes into a test log.
Tools we use
We use Meta Ads Manager A/B testing, Google Ads experiments for campaign-level changes, and ad variations or asset reporting where appropriate. Landing page tests can be run through our own page builds with traffic split and tracking in GA4. Results are summarised in Looker Studio or a shared test log.
What affects results, timeline and cost
The main constraint is data volume. Accounts with few conversions per week need longer tests or bigger differences between variants to reach a clear result. Small differences on small budgets may never become conclusive, which is why we prioritise bold tests before fine-tuning.
A single test often runs for a couple of weeks or more, depending on volume. Testing is usually part of ongoing management rather than a separate project. We never promise that a test will produce a winner; sometimes the useful result is learning that a change makes no difference.
Optimisation beyond tests
Not every improvement needs a formal test. Pausing ads with clearly poor results, adding negative keywords, fixing broken tracking or refreshing tired creative are routine optimisation tasks. We keep formal tests for decisions where the answer is genuinely uncertain and worth knowing.
Common testing mistakes
- Changing creative, audience and landing page at the same time, then guessing which one mattered.
- Ending tests after a day or two because one variant is ahead.
- Testing tiny details, like button colour, on low traffic.
- Judging on click-through rate when the goal is qualified leads or sales.
- Not recording results, so the same test is repeated months later.
Frequently asked questions
How long should an ad test run?
Long enough to collect a meaningful number of conversions and cover full weekly cycles. The exact duration depends on your budget and conversion volume.
Can small budgets run tests?
Yes, but tests should focus on big differences, such as a completely different angle or offer, because small tweaks need more data to measure.
Do platform algorithms make A/B testing unnecessary?
No. Platforms optimise among the options you give them, but they cannot tell you which strategic choices, such as offers or messaging, work best without structured comparisons.
Will I see the test results?
Yes. Each test is recorded with its hypothesis, setup, result and decision, and summarised in your reports.
What do you do when a test has no clear winner?
We record it as a result, because learning that a change makes no difference is useful. Then we either keep the simpler version or move on to a bolder test with a clearer difference between variants.
Talk to us about a/b testing & optimization
Structured tests on creative, audiences and pages, one variable at a time.