Testing an ad creative helps you understand which message and design attract suitable customers. To get useful results, define in advance what you are testing, how you will compare variants and which metric will guide the decision.
If you change the offer, image, audience and landing page simultaneously, you will see a difference in results but will struggle to explain its cause. Start with a specific question and comparable conditions.
For a service business, the primary metric can be the cost per qualified lead: an enquiry that meets predefined requirements. For example, the service, geographic coverage and project budget suit the potential customer.
For an online shop, assess purchases and their economics, accounting for returns and margins. If sales data is still limited, intermediate actions can indicate a direction, but conclusions about sales will remain provisional.
CTR, the share of impressions that lead to clicks, helps diagnose how an ad attracts attention. However, a high CTR can accompany irrelevant enquiries. Separate the following in advance:
A useful template:
If we change [a specific element] for [an audience], then [the primary metric] will improve because [an explanation].
A hypothetical example for a website development service: “If we show what the project includes instead of making a general promise of attractive design, visitors will understand the offer more clearly and the share of qualified leads will increase.”
Prepare two ad versions to test this. Keep the current emphasis in one and introduce the new emphasis in the other. Retain the other important conditions: the audience, landing page, offer and lead tracking method.
You can initially compare entire concepts to select a promising direction. This test measures the result of the overall presentation. To understand the effect of a particular element, run a subsequent test with more similar variants.
You can also test changes to the offer or audience, but identify these as separate objectives. This prevents results from being incorrectly attributed solely to the banner design.
If the advertising platform provides a suitable experiment tool, use it for a controlled comparison. Ads simply running alongside one another may receive different impression volumes and reach different audiences. Automated allocation does not necessarily create equal test conditions.
In its guidance for Search ads, Google recommends limiting the number of elements changed, recording the timing and collecting enough data before drawing conclusions. Ad variations and experiments are available for Search advertising. Google's guidance on testing creative messages.
Before launch, agree on:
Check that forms work, enquiries reach your CRM or another agreed channel, and the lead source is recorded.
The following table contains invented data to illustrate the evaluation method.
Variant A generates cheaper enquiries, while variant B produces a lower cost per qualified lead. If the business prioritises suitable customers, evaluating only the overall cost per enquiry will lead to a different choice.
However, six and eight qualified leads are small volumes. The table suggests a preliminary direction but does not prove a lasting advantage. Check whether conditions are comparable, wait for enquiries to be processed and continue collecting data if necessary.
Choose one important hypothesis and a small number of variants. The more versions share the same limited budget, the less information you collect about each.
Avoid a universal rule such as “three days are enough for every test”. The required data volume depends on how often the target action occurs, the difference between variants and the level of confidence needed.
If data is insufficient, record the uncertainty. Running variants sequentially in different weeks can provide an indication, provided you account for changes in demand, competition and delivery conditions.
For every test, record:
This helps the team develop successful directions and understand why a particular decision was made.
We can prepare variants through our advertising banner and creative design service. Before ordering, agree on the hypothesis matrix and required formats. The scope of work is set out in our commercial proposal. Include experiment setup and result analysis separately in the advertising work plan.
A useful test ends with a clear decision: what to retain, what to change and which data is needed for the next step.