The short answer: test one idea at a time, run each ad long enough for the numbers to mean something and write down the rules for pausing an ad before you launch. On a small budget, discipline matters more than volume.
The temptation is to launch ten versions, check them every morning and pause the ones that look weak by lunchtime. This is how you end up with a "winner" that was only lucky.
Start with a question, not a design
Every test should answer a question that could change what you do next.
- Does proof from a customer beat an explanation of the product?
- Does a video beat a set of photos?
- Does naming the audience in the first line bring better enquiries?
- Does a message about timing, such as "before the season", beat a message about quality?
If you cannot say what you will do differently depending on the answer, do not run the test.
Change one thing at a time
If one ad has a different video, headline and text from another, you cannot tell which change mattered. Keep the rest the same.
On a small budget, three to five ads is enough for a first round. Make them different enough to teach you something: a testimonial, a product demonstration, an urgency message, a set of photos.
Write the rules before launch
- Do not judge early. Give each ad time and enough spend before you compare. The first day is mostly noise.
- Define the measure. Cost per qualified enquiry, or cost per conversation if you must, and name which.
- Define the review date. One date, not a daily check.
- Define what you pause. For example, the two ads that are clearly weakest at the review.
Rules written in advance protect you from reacting to a good or bad morning.
Learn from the losers too
When an ad fails, ask why. Was the promise unclear? Was the first line weak? Did it attract the wrong people? Write it down. Over several rounds, these notes become the most valuable thing the account owns.
What to test next
- The idea that won: try a new version with a different opening line.
- The format: the same message as a video, a carousel and a single image.
- The audience: the same ads to a different segment.
Test in that order. A strong idea carries over to any audience, but a weak one does not become strong by changing who sees it.
An illustration: a first round of four ads
Imagine four ads for the same offer: a customer testimonial video, a product demonstration video, a carousel of photos and a single image with an urgency message. Each is shown to the same audience at the same time.
After the review date the numbers might show that the testimonial and the carousel brought the most qualified enquiries, while the single image brought many conversations that did not qualify. The next round keeps the testimonial and the carousel and tests two new opening lines on each. In two rounds you have learned which kind of proof your buyers respond to.
Keep a testing log
Write one line per test: the idea, what you changed, what you expected, what happened and what you will do next. The log is the memory of the account. When someone new joins the team, it saves months.
A one-line brief for each test
Write each test in one line before you launch: "We think [change] will improve [measure] for [audience], because [reason]. We will judge it on [date]." If you cannot complete the sentence, the test is not ready. The line also stops arguments later, because everyone agreed what a good result looked like before the numbers arrived.
Keep the lines in the testing log, and read them again before you plan the next round.
Where good ideas for tests come from
Testing is only as good as the ideas being tested. Waiting for inspiration is slow, so use sources you already have.
- The questions customers ask. The first questions in messages and calls reveal what people want to know before they decide. Each one can become an ad angle.
- The objections the sales team hears. An ad that answers a common objection can stop the same objection later in the conversation.
- Reviews and testimonials. The words real customers use are often better than the words the business would choose.
- Competitor ads. Public ad libraries show what others are saying. Look for what they all say, and then test something different, not a copy.
- Your own past results. Which posts or ads got the most replies, and what did they share?
Write each idea as a short sentence in the testing log. Rank them by how much you expect them to teach you, and test the top ones first.
Reading small numbers honestly
A small budget produces small numbers, and small numbers are easy to misread. Three habits help.
- Do not declare a winner from a handful of results. If one ad has a few more enquiries than another, the difference may be chance. Wait until the gap is large and has lasted.
- Look at the whole path, not just the click. An ad that gets many clicks but no useful enquiries has not won.
- Respect delivery effects. The platform may spend more on the ad it prefers early on, which makes a fair comparison harder. Where possible, give each variant a comparable chance.
When the evidence is thin, say so in the log and keep testing. An honest "not sure yet" is more useful than a confident wrong answer.
A weekly testing routine that takes an hour
- Review last week's test against the question you wrote. Record what you learned in one or two sentences.
- Decide what to keep, retire or iterate and note the reason.
- Choose the next test from the ranked list and write its one-line brief.
- Prepare the new variant, changing only the element being tested.
- Launch and check delivery after a day to be sure that both variants are running.
Done every week, this builds a record of what works for your audience. Within a few months, that record is a real asset: it tells you how to write, what to show and who to speak to, and it is yours to keep.
How we do it
We plan a test round, set the rules and review once, then keep the winners running. If you want a second pair of eyes on a campaign that is not behaving, tell us what you are running.