Skip to content
GullySales

Notes for owners · Digital marketing

How to test advertising creative

Digital creative can be tested after it runs. A hoarding, a newspaper ad and a radio spot have to be tested before they are made, and the tests cost nothing.

The GullySales team · Updated 21 Sept 2026 · 8 min read

Creative testing splits into two jobs that get confused with each other. Digital creative is tested after it goes live, by spending real money against real buyers and counting enquiries. Print, outdoor, radio and cinema cannot be tested that way, because one version gets printed and there is no second chance, so they are tested before they are made, by hand, in a room. The offline tests take an afternoon and cost nothing, and skipping them is how a hoarding nobody can read stays up for three months.

The tests to run before anything is printed

The three-second test. Put the hoarding artwork on a screen, show somebody who has not seen it for three seconds, take it away, and ask what it was for and whose it was. If they cannot name the business, the name is too small or the picture is too busy.

The squint test. Look at the layout with your eyes half shut. Whatever survives is what a reader will actually take in. If the only thing left is a photograph of a building, the ad is about a building.

Cover the logo. Would a competitor's logo sit there just as comfortably? If yes, you have made a category advertisement and any of your rivals could pay for it.

Read it aloud with a stopwatch. For radio, every time. Most scripts approved on paper run four seconds long, and the studio decides what to cut at ten at night.

Print it at actual size. Tape the newspaper ad, at its real column width, into yesterday's paper and look at it on the table where people read. Type that looked confident on a laptop vanishes at 20 per cent of that size.

Hold the phone at arm's length, sound off. For reels and feed video, this is the only honest preview. The frame your designer loves is usually the one with the platform's buttons over it.

Ask one person on the shop floor. They hear the actual questions buyers ask. They will tell you in ten seconds that nobody asks about what your headline is about.

When you show it to real buyers

Ten people who look like your buyers are enough for a small business, and you do not need a research agency to do it. What matters is the question you ask.

Do not ask which version they prefer. People answer that politely and wrongly. Ask what the advertisement was about, what it costs, what they would have to do next, and whether anything confused them. Then ask them again after five minutes of talking about something else, because recall is the thing you are buying on most media.

Watch what they read first, and where their eyes stop. That tells you more than any answer.

What a digital test has to get right

RuleWhy
One variable at a timeTwo changes and you cannot say which one did it
The platform's split test, not three ads in one setOtherwise the platform, not you, decides what gets the budget
A week at minimum, a fortnight preferredWeekdays and weekends behave differently in most trades
Judged on enquiries, better still on qualified onesThe best clicked ad regularly brings the worst enquiries
Enough events to be believableTen conversions split five and five is a coin toss, not a result
The same landing page for both versionsChange one thing, not two

The last row is the one that gets broken most. A team tests two creatives and sends them to two different pages, learns nothing, and repeats it next month.

Test the offer before you test the artwork

The order matters because the size of the effect differs so much. The offer first: free consultation against a fixed price, this week against no deadline, an amount off against something included. Then the picture, then the headline, then the body copy.

For example, a dental clinic running two versions of the same ad, one offering a free consultation and one naming a fixed price for a cleaning, will see a difference in both the number of enquiries and in who turns up. That is a real finding. Two fonts is not.

Testing offline, where it is possible

Radio is tested by station and by script. Run script A on one station and script B on another with a different number on each, for the same fortnight, and compare the calls. The stations differ, so it is rough, and it is still better than a guess.

Newspapers are tested by edition, since the same paper prints different city and district editions on the same morning. Two versions, two numbers, one day.

Hoardings cannot be tested. One creative goes up for a cycle and that is it, which is exactly why the three-second test and the printed proof matter on that medium more than anywhere else.

And on every offline medium the counting depends on a number or a code that appears nowhere else, which is what call tracking is for. Without that, every finding is an opinion.

What testing cannot tell you

Whether the brand line works, because that takes a year and shows up in what people call you rather than in a dashboard. Whether a campaign would have done better with a different medium, since you only ran one. And anything at all from a week that contained Diwali, an exam result, or a long weekend, because the week did the work and not the creative.

Say that out loud at the review. A test that ran during a festival week gets repeated in a plain week before anybody acts on it.

What to do next

Take the creative sitting in your approval folder now, and run the three-second test and the cover-the-logo test on it this afternoon. Then decide, for the digital part, the one thing you are testing and how many enquiries you will need before you call it. Write both down before the campaign starts.

We run these checks on client creative before it goes to print or to a platform, and the campaign analytics and performance reporting is set up so a version can be traced to an enquiry rather than to a click. If your current campaigns cannot be traced that far, that is what the free audit looks for first.

Questions

Questions owners ask.

How long should we run a creative test?
Long enough to collect enough enquiries to tell the versions apart, and never less than a full week. A week covers the difference between a Tuesday and a Sunday, which is large for most businesses. If a fortnight has produced eleven enquiries in total, you have not run a test, you have run an advertisement.
One ad set with three creatives, or three ad sets?
Use the platform's own split test, which divides the audience so the same person does not see two versions. Three creatives inside one ad set is not a test, because the platform quietly spends most of the money on whichever one gets early clicks, and early clicks are not orders.
Can a newspaper advertisement be tested at all?
In a limited way. Papers sell separate city and district editions, so two versions with two phone numbers can run on the same day in two editions and be counted. What you cannot do is change your mind after the file has gone to press, which is why the desk tests before printing matter more here.
What should we test first?
The offer, then the picture, then the headline. The offer moves results far more than anything the designer does, and it is the cheapest thing to change. Businesses that test button colours while the offer is wrong are polishing the fourth most important thing.
Can we ask ten customers which version they prefer?
Ask them what they remember and what they think it costs, not which one they like. People are poor at predicting what will make them act and polite about telling you. What they can report accurately is what they took away five minutes later.

Get a free audit of how you sell, and a scored report of where the work is.

Book a free audit