Facebook ad creative that works
Hooks, offers and formats that stop the scroll in Malta.
Facebook & Instagram ads · A/B testing
Facebook ads A/B testing,
without wasting budget.
A Facebook ads A/B test shows two versions of one thing to separate groups, so nobody sees both, and tells you which costs less per result. Test one change at a time for at least 7 days. Malta Digital runs these tests inside Meta ads management, from €1,249 a month plus ad spend.
One variable7 to 30 days, same budget
Short answer
Pick one variable, such as the image or the audience. Use Meta’s A/B test tool instead of switching ads on and off. Give each version the same budget and run it for 7 to 30 days. Then judge on cost per result. Malta Digital does this for clients as a Facebook ads agency in Malta.
Key takeaways
- How long should a Facebook ads A/B test run? Meta recommends at least 7 days and allows up to 30. Run it longer if your customers usually take more than a week to buy or enquire.
- How much budget do I need for an A/B test? Enough for each version to get a fair number of results. Our working rule is about 30 per version. Use the free planner on this page with your own cost per result.
- Can I test more than one thing at once? You can with a custom test, but Meta suggests one variable for the clearest answer. If you change the image and the audience together, you will not know which made the difference.
The basics
What a real A/B test is.
A/B testing compares two versions of an ad strategy. Meta shows each version to a separate part of your audience and makes sure nobody sees both. Then it tells you which one performed better (Meta Business Help: About A/B testing, opened 8 Nov 2026).
Example sceneA Gozo farmhouse rental has two hooks. One says “Private pool, 10 minutes from Mġarr”. The other leads with “Sleeps 8, book direct”. The owner swaps them every few days and cannot tell which works. With an A/B test, each hook gets the same budget and its own half of the audience for 14 days. At the end there is one cost per booking enquiry for each. This is an example, not a client result.
That is different from running two ads side by side in one ad set. There, Meta pushes budget to whichever ad it likes early. One ad may get most of the spend before you learn anything.
Meta also warns against testing by turning ad sets on and off by hand. Audiences overlap and the results become unreliable. A proper test splits the audience evenly.
This guide is for owners paying for Meta ads in Malta who want to know what is actually working, not guess.
Free tool
Plan your test with the free planner.
Type the daily budget for the whole test, your usual cost per result and the number of days. The planner shows how many results each version should get. Nothing leaves this page.
A/B test planner
Meta allows 1 to 30 days and recommends at least 7, read 8 November 2026. The budget is split evenly.
Your test plan
This planner needs JavaScript. The maths: results per version = daily budget ÷ 2 × days ÷ cost per result.
Example only: €40 a day for 14 days at €9.68 a lead gives each version about 29 leads. That is close to our working rule of 30 per version. €20 a day for 7 days gives about 7 each, which is too few to call.
What to test
What to test first.
Meta lets you test creative, audience, placement or a custom mix (Meta Business Help: Selecting a variable for your A/B test, opened 8 Nov 2026). Meta suggests starting with one variable for the clearest answer. Here is the order we usually follow for Malta businesses.
| Order | Variable | Example test | Level in Ads Manager |
|---|---|---|---|
| 1 | Offer or hook (creative) | “Free quote in 24 hours” vs “Prices from €X” | Ad |
| 2 | Format (creative) | 15-second video vs a single image | Ad |
| 3 | Audience | All of Malta vs a 10 km radius | Ad set |
| 4 | Optimisation goal | Leads vs landing page views | Ad set |
| 5 | Placements | Advantage+ placements vs Feed only | Ad set |
| 6 | Landing path | Instant form vs your own landing page | Ad |
Start with the offer. It moves cost per result more than colours or button text. Write a clear hypothesis first. Meta’s own example is “my cost per result will be lower when I optimise for landing page views”.
For ideas, see our guides to Facebook ad creative and landing pages for Facebook ads.
Length and budget
How long and how much.
Meta recommends tests of at least 7 days. Tests can run from 1 to 30 days, but shorter than 7 may be inconclusive (Meta Business Help: Best practices for A/B testing, opened 8 Nov 2026). If your customers take longer than a week to decide, run it longer.
Meta also says the budget should produce enough results to pick a winner with confidence. It recommends the same budget for both versions.
Two more rules from Meta matter in Malta. Use an audience large enough for the test. And do not use the same audience in another campaign at the same time, because overlap can spoil the result.
Our working rule is about 30 results per version before we call a winner. That is our habit, not a Meta number. Meta shows its own confidence figure in the test report.
Proof
Proof: testing inside a disciplined account.
From 1 January to 28 May 2026, a Facebook and Instagram account we managed spent €95,470.88 and tracked €573,222.05 in purchase value from 9,085 purchases. Average ROAS was 6.00. Creative was tested inside clear sales campaigns, and budget moved to what won.
Tracked purchase value is not profit. The point is the method: test one thing, keep the winner, test the next thing. Small gains add up over months.
Want a testing plan for your account? Apply now for Meta ads testing
Not sure what to test first?
Apply now and tell us your goal and current ads. We will suggest the first test that could move your cost per result.
Free tool
Read your result.
When a test ends, enter the spend and results for each version. The reader shows the cost per result for each and how far apart they are. Meta’s test report is still the official answer.
A/B result reader
Your numbers in, cost per result out. Nothing leaves this page.
Your result
This reader needs JavaScript. Cost per result = spend ÷ results, for each version.
Meta picks the winner by comparing cost per result. It simulates the outcome tens of thousands of times and gives a confidence percentage (Meta Business Help: How winners are determined in A/B tests, opened 8 Nov 2026). Meta says a “winner” can still cost more if the test was too short or had too few results.
Malta
Why testing pays off in Malta.
Malta feeds are crowded with local ads. In 2024, 60.4% of Malta firms with 10 or more staff paid to advertise online. The EU figure was 32.6%. And 36.4% of those Malta firms paid for ads targeted by location, against 14.8% in the EU (Eurostat isoc_cismt (paid online ads, 2024), opened 8 Nov 2026).
When many nearby firms target the same people, small edges matter. A better hook that cuts cost per lead by a fifth gives you a quarter more leads for the same spend.
Malta is also small, with 588,254 residents on 1 January 2026 (Eurostat tps00001 (population on 1 January), opened 8 Nov 2026). Audience tests on tiny slices rarely get enough results. Test creative and offers first, on a broad Malta audience.
Want each euro of your test to teach you something? Apply now for Meta ads testing
Mistakes
A/B testing mistakes we fix most often.
Testing three things at once. New image, new text and new audience. You cannot tell which one won.
Stopping after two days. Meta says tests under 7 days may be inconclusive.
Uneven budgets. Give both versions the same money.
Testing button colours first. Test the offer. It usually matters most.
Not rolling out the winner. A test only pays when you use what it found.
Each new test also starts a fresh learning week. Read our guide to the Facebook ads learning phase to see why. For costs, see Facebook ads cost in Malta.
A few straight answers
What you might be wondering.
How long should a Facebook ads A/B test run?
Meta recommends at least 7 days and allows up to 30. Run it longer if your customers usually take more than a week to buy or enquire.
How much budget do I need for an A/B test?
Enough for each version to get a fair number of results. Our working rule is about 30 per version. Use the free planner on this page with your own cost per result.
Can I test more than one thing at once?
You can with a custom test, but Meta suggests one variable for the clearest answer. If you change the image and the audience together, you will not know which made the difference.
How does Meta pick the winner?
Meta compares the cost per result of each version. It simulates the outcome many times and gives a confidence percentage.
Is an A/B test the same as two ads in one ad set?
No. In one ad set, Meta pushes budget to the ad it prefers early. An A/B test splits the audience so nobody sees both, and gives each version a fair share.
What should a Malta business test first?
The offer or hook. It usually moves cost per result the most. Test format next, then audience and goal.
Does Malta Digital run A/B tests for clients?
Yes. Testing is part of Meta ads management from €1,249 a month plus ad spend. We plan one clear test at a time and report what won.
Sources
Where these facts come from.
- Meta Business Help: About A/B testing, opened 8 Nov 2026
- Meta Business Help: Best practices for A/B testing, opened 8 Nov 2026
- Meta Business Help: How winners are determined in A/B tests, opened 8 Nov 2026
- Meta Business Help: Selecting a variable for your A/B test, opened 8 Nov 2026
- Eurostat isoc_cismt (paid online ads, 2024), via the Eurostat API, opened 8 Nov 2026
- Eurostat tps00001 (population on 1 January), via the Eurostat API, opened 8 Nov 2026
Keep reading
Related guides.
Facebook ads learning phase
The 50-results rule, edits that restart it, and a free checker.
Facebook ads metrics explained
Which numbers matter and which ones to ignore.
A useful first step
Want tests that find winners, not noise?
We plan one clear test at a time, run it the way Meta recommends and roll out what wins. Meta ads management from €1,249 a month plus ad spend, with the ad account in your name.
Step 1 of 2
Apply now for Meta ads testing
A few quick questions, about a minute. We usually reply within 2–3 hours.