Creative testing for clothing ads: change one thing at a time

Clothing ads usually change many things at once: a new model, a new location, new text and a new garment. When one ad wins, you can’t tell whether buyers wanted the garment or just liked the photo. For a style test, everything except the garment should be the same in every ad.

This is a style test. Testing photos, hooks or ad copy is useful too, but run it separately, with the garment fixed.

Meta’s A/B testing lets you test which creative, audience, placement or objective works best ^[1]. Its guidance is that only one variable should change while the rest of the campaign setup and creative stays identical, and you can test up to five versions ^[1]. Meta says it splits the audience into random, non-overlapping groups so the results are comparable ^[1].

Meta also has a creative testing option in Ads Manager, with its own help page ^[2]. How it is set up has changed more than once, so follow Meta’s page and what Ads Manager shows you rather than a third-party summary.

Fewer than you think. Every version takes a share of your budget, and each one needs enough views and actions to compare. Meta’s A/B tests allow up to five versions ^[1]; Shopify’s guide to creative testing suggests considering three to four versions at once ^[3].

• Start with your bestseller and two or three new versions.
• Drop the weakest ones and test new versions against the strongest.
• Don’t add a version halfway through: it starts behind the others.
• Set your stopping rule before you start, and use it for every version. Shopify’s guide to creative testing calls at least a week a good rule of thumb ^[3]; its guide to validating product ideas suggests running small tests until you log 100 conversion events such as orders or sign-ups ^[6].

When results come in, you need to know which version each number belongs to, in your ad reports and in your store’s sales. Use one name per version, the same everywhere: the image file, the ad, and the link’s UTM tags.

Google Analytics says to always use utm_source, utm_medium and utm_campaign, and describes utm_content as the tag that tells creatives or links apart ^[5]. Values are case-sensitive, so utm_source=google and utm_source=Google count as different sources, and Google suggests using lowercase ^[5]. A full link for version B could end in ?utm_source=meta&utm_medium=paid_social&utm_campaign=test-linenshirt&utm_content=test-linenshirt-b.

• Never reuse a letter for a different version, even after you drop one. Old reports would then point at the wrong garment.
• Put the shape in the file name too (for example TEST-LINENSHIRT-B-4x5), so the right size goes to the right placement. See ad sizes for clothing ads.

Clicks tell you an image caught someone’s eye. For a style test you want to know whether people would buy it, so look further down the funnel:

• Purchases or pre-orders per version, where you take them. The founder of one T-shirt brand Shopify interviewed said conversions, meaning purchases, are what he looks at ^[4].
• Add-to-carts and checkouts started, if you don’t sell yet.
• Sign-ups to a waitlist for each version.
• Comments that ask when, how much or in what size.

Shopify’s guide to creative testing lists the usual metrics: impressions, engagement, conversions, watch time, time on page and click-through rate ^[3]. Pick one main measure before you start, so the result isn’t whichever number looks best afterwards.

1. Write down what changed in the winning version, and the result you based the decision on.
2. Send your factory the change. See tech pack or spec sheet.
3. If you took pre-orders, keep the ship date you gave. See pre-orders for clothing.
4. Replace the concept images with real photos once the garment exists. Until then, label them. See AI labels on ads and product images.
5. Plan the next test from what you learned. Our guide to testing a new style of your bestseller walks through the whole loop.