How-to
How to run a simple ad creative test
Pick one question, test two to four genuinely different concepts against each other with the same audience, budget and placement, decide in advance which result you will judge by, and run it long enough to see more than a handful of results. Then keep the winner and test a new idea against it.
The method in six steps
- Ask one question. For example: does a price-led ad or a quality-led ad bring more enquiries? A test that tries to answer five things answers none.
- Make the concepts genuinely different. Two to four ideas that differ in the thing you are testing, such as the benefit or the offer. Keep everything else the same. A colour change is a variation, not a test of an idea.
- Keep conditions fair. Same audience, same budget per ad, same placements, same start date and same landing page.
- Pick the measure first. Choose what counts as a win before you start, ideally cost per enquiry or booking rather than clicks alone. Clicks are easier to get than customers.
- Run it long enough. Leave it for at least a week or two so weekdays and weekends average out, and until each concept has more than a handful of results. A couple of enquiries either way is noise, not a verdict.
- Decide and record. Keep the winner, stop the clear loser, and write down what you learned. Then test a new concept against the winner.
What to leave alone during a test
- Do not edit the ads, audience or budget partway through. Each change restarts learning and muddies the result.
- Do not stop an ad after a day because it looks poor. Early numbers swing a lot.
- Do not compare results from different weeks, audiences or offers.
Small budgets
If you only spend a little, test fewer ideas for longer. One clear question every few weeks beats many ideas starved of spend. The concepts guide shows a way to work out how many you can afford.
Using the tools
Meta Ads Manager has a built-in A/B test feature that splits an audience evenly between versions, so the groups do not overlap. Menu names change, so check Meta's current documentation. Running several ads in one ad set also works for simple comparisons, but delivery is not split evenly, so read the results with care.
Keep a simple log
For each test record: the question, the concepts, dates, spend, results, the decision and what you will try next. After a few months this log is more useful than any benchmark, because it describes your customers.
Where the concepts come from
A creative strategy gives every concept a job, so each test is about a real difference. See the creative subscription.