How to plan a paid social creative test

5 steps to a successful testing plan

Testing is key to successful digital marketing. However, poorly thought out testing won’t get you anywhere. Read on for our 5 steps to plan creative tests for paid social.

1. Define success

Start by defining your goal, a KPI to measure it, & a target. For example:

Every successful campaign starts by defining what you’re trying to achieve, & keeping that at the centre of the decision-making process.

In the context of testing, your goal will help you decide what you want to test, & decide which variant is the winner.

It’s also important to consider whether you will have a high enough volume of results to determine a winner. For example, if you only generate on average 3 purchases a week through your campaigns, & you want to run your test for 2 weeks, then you won’t get enough data points to determine your winner based on ROAS. In this case, you should choose a higher volume KPI, like clicks & CPC (cost per click).

2. Decide what you’ll test

You need to be testing things that are impactful enough to actually make a difference. Realistically, changing the colour of the font isn’t going to give you significant results, so don’t waste your time on small tests like this.

A good place to start is by asking yourself what are the different reasons a user might perform your desired action. In our example, this would be a purchase.

Imagine you’re a clothing brand. Some reasons why a person would buy clothes from you would be:

Think as top line as possible. If you need help finding reasons, generative AI like ChatGPT can provide a great starting point.

Each of these can be broken down further. For example, there are many different impressions a user might want to create of themselves by buying clothes:

We can then break each of these down further, by thinking of different ways that we can convey to the user that they can achieve this need by buying your clothes in an ad:

Going through this process, you can easily build multiple unique ad ideas, answering different user needs, & communicating this to them in a variety of ways.

Rather than customer needs, you could try different starting points, like different areas of your business/charity, or unique selling points, & work down from there.

3. Choose a testing method

You could either A/B test 2 variants under identical conditions, or you could launch multiple variants at the same time & see which the platform’s AI machine learning algorithms favour. Both are valid methods, & have pros & cons:

Multivariate is usually the right choice if you have smaller budgets, a lot of different ad ideas, or need to optimise towards the best variants quickly, while A/B is better for if you have a specific question you want reliable answers to e.g. “Which gets a better ROAS, ads advertising our clothes as cool or professional?”

Another way of thinking about it, is multivariate is best if you just want the best creatives spending quickly, while A/B is better if you want to know why they are the best.

4. Choose your budget

The higher the budget, the more reliable your results will be. 

For A/B tests, aim to spend at least enough to generate at least 20 of your desired action over the lifetime of the test. 

In our example, if your usual CPA (cost per action/purchase) is £50, you should spend at least £1000. If this is too high, then you could decide to judge the winner based on CPC instead. Since it’s much cheaper to generate clicks, you could run the test on a far smaller budget e.g. £50.

With multivariate testing, you can add the creatives you’re testing into your usual campaign structure alongside currently live ads, so there’s no need to allocate a specific amount of budget to them.

In general, you should aim to use 5-10% of your overall advertising budget for testing.

Note: These are just guidelines, & the amount of spend you need to achieve statistical significance could vary wildly.

5. Set a duration

We recommend running an A/B test for between a week & a month.

For multivariate testing, as long as your creatives aren’t harming overall performance, you could leave them live for a couple of months & allow the platform to naturally optimise towards/away from them.

Do good, better

If you’re looking for a digital marketing agency to help you achieve growth through testing, get in touch!

How to plan a paid social creative test

Oxford educated, Gwenno’s expertise now spans Paid Search, Social & Display, working on a mix of ecommerce, charity, event, education & B2B clients. In addition to tackling creative strategy, she is also the agency's tracking expert and oversees data & reporting projects.