How to test a new style of your bestseller before you make it
Last checked Oct 22, 2026 · Reviewed by Museify team · 8 min read
Start from your bestseller, because your buyers have already told you they like it. Pick a few new versions that each change one thing (a colour, a print, an occasion or one detail), show every version on the same model, pose and light, and put them in front of buyers: ask your audience, run a small ad test, or open a pre-order with a real ship date. Shopify’s guide compares pre-orders with unique landing page visitors and says the pass rate can range from 10% to 20% [1], and Meta’s own testing advice is to change only one variable at a time [3]. These are rules of thumb, not promises. The images are concept images; when a version wins, send your factory what changed, get a sample made and swap in real photos.
On this page
Why start from what already sells
A brand-new design asks two questions at once: do people want this kind of garment, and do they want this version of it. A new style of your bestseller asks only the second. The cut, fabric and fit have already sold to your buyers; what you are testing is whether a new colour, print or detail makes it sell to more of them, or to different ones.
- You already know the garment can be made, at what cost, and by whom.
- Your buyers already trust the fit and the fabric, so a test result is mostly about the change you made.
- Your factory already has the pattern, so a winning version usually means a change to an existing product rather than a new one.
Use your own bestseller and your own photo of it. Testing a version of someone else’s product, or using another brand’s photos, is a different thing and not what this guide covers.
Pick variants that change one thing
Each version should differ from your bestseller in one way only. Keep everything else locked: the category, cut, length, fabric and fit. Your bestseller itself stays in the test as version A, the baseline the others are compared against.
| Kind of change | Example | What it tells you |
|---|---|---|
| Colour | The same shirt in butter yellow | Whether a new colourway is worth a production run. See testing a new colourway. |
| Pattern | The same dress in a small floral print | Whether a print adds appeal or narrows it |
| Occasion | The same dress styled for a wedding guest instead of the beach | Whether the garment sells to a new moment, with no change to the product itself |
| One detail | A square neckline instead of a round one | Whether a design change is worth a new sample. Treat it as a concept until a sample exists. |
Keep the number of versions small. Every version splits your audience and your budget, so a test with three or four versions gives each one a fairer share than a test with ten. Our guide to creative testing for clothing ads goes into how many to run.
Make it a fair test: same model, pose and light
If version B is on a different model, in a different place or in different light, a better result for B could be the photo rather than the garment. Meta’s advice for its own A/B tests is the same idea: only one variable should change, while the rest of the campaign setup and creative stays identical [3]. Its examples of creative elements to test include product colour [3].
- Show every version, including your bestseller, on the same model, in the same pose, light and framing.
- Use the same ad text, headline and landing page for every version.
- Make each ad size natively for each version, so no version loses its hem or shoulders to a crop. See ad sizes for clothing ads.
- Check every image against the real garment before it runs. Our guide to putting your clothes on an AI model lists what AI tends to get wrong.
Three ways to test: ask, advertise, pre-order
There is no single right way. Most sellers start with the cheapest signal and move on to stronger ones for the versions that survive.
| Method | What you do | What to look at |
|---|---|---|
| Ask your audience | Post the versions to your followers or email list and ask which one they would buy. Shopify’s guide notes that your followers will tell you what they want to see [1]. | Replies and votes from people who have bought from you before. Likes alone are a weak signal. |
| A small ad test | Run the versions as ads with a small daily budget. The founder of one T-shirt brand Shopify interviewed tested at about $5 a day and judged a design likely to fail if it got no bites within three to four days [2]. | Purchases, add-to-carts or sign-ups rather than clicks; that founder said conversions are what he looks at [2]. Shopify suggests running small tests until you log 100 conversion events such as orders or sign-ups [1]. |
| A pre-order or coming-soon page | Put the version on a product page with a real ship date, or on a password-protected Coming Soon page that collects pre-orders, which Shopify suggests [1]. | Pre-orders compared with unique landing page visitors; Shopify says the pass rate can range from 10% to 20% [1]. For an email waitlist, it suggests aiming to convert up to 5% of subscribers [1]. |
Selling something that doesn’t exist yet: the rules
Showing a new version to measure interest is one thing; taking orders for it is another. In the US, the FTC’s Mail, Internet, or Telephone Order Rule says you must have a reasonable basis for any shipping time you state, and if you state none, for shipping within 30 days [4].
If you want to take sign-ups for a style you may never make, the FTC describes a narrow route it calls dry-testing, with strict conditions: you make no suggestion that the item will be shipped, you disclose that it is only planned and may not ship, you tell people promptly if you decide not to go ahead and refund any payment, and you don’t send anything else instead [4]. If those conditions aren’t met, the rule applies [4].
- Our guide to pre-orders for clothing covers ship dates, delay notices, refunds and EU and UK rules.
- Label the images as AI-generated concept images wherever buyers see them. Our guide to AI labels on ads and product images covers what Meta, TikTok and Google ask for.
- Our guide on using AI models for your clothing brand covers what AI model photos can and can’t show.
After a winner: from concept images to a sample
- Mark the winner and write down what changed from your bestseller, in one line.
- Send your factory that change, the colour you want and anything about the fabric. A one-page spec sheet is enough to start the conversation; a full tech pack comes before sampling. See tech pack or spec sheet.
- Check the first sample against your bestseller itself, not against the AI image, since small details in the image may be off.
- If you took pre-orders, keep buyers updated against the ship date you gave them.
- Once the garment exists, replace the concept images with real photos.
Every step, in more detail
- Can I use AI models for my clothing brand?: what AI model photos are good for, and how to tell buyers.
- How to put your clothes on an AI model from one photo: which photo to start from and what to check.
- How to test a new colorway without a photoshoot: changing only the colour, and screen colour versus fabric.
- Ad sizes for clothing ads: 1:1, 4:5 and 9:16: what each Meta and TikTok placement wants, and safe zones.
- Creative testing for clothing ads: how many versions, naming and UTM tags.
- AI labels on ads and product images: Meta, TikTok, Google and the EU.
- Pre-orders for clothing: ship dates, delays and refunds.
- Tech pack or spec sheet: what to send your factory after a style wins.
FAQ
Do I need a sample before I can test a new style?
Not to measure interest. Concept images let you see whether buyers respond to a version before you pay for a sample. You do need a sample before you promise fit, fabric or details to a buyer, and before a production run.
How long should a test run?
Long enough for each version to get a fair number of views and actions. The sources give different rules of thumb, from a few days for a small ad test [2] to running until you have 100 conversion events [1]. Decide your rule before you start, and apply it to every version the same way.
What if my bestseller wins against every new version?
That is a useful result too: it tells you to restock the original rather than add a version. Try a different kind of change in the next test.
Can I test a version of a garment I don’t sell yet?
This guide is about new versions of your own bestseller. If you’re starting from an idea instead, the same rules apply: change one thing at a time, label concept images, and give a real ship date if you take orders.
How Museify helps
Upload a photo of your bestseller (on a model, flat lay, mannequin or hanger) and tick that it is your product; the person in the photo is never edited, and new images use one of our eight AI models. Your bestseller is version A, and up to five more versions each change one thing (a colour, a pattern, an occasion or one detail), all on the same model, pose and light. Each version is made natively in 1:1, 4:5 and 9:16; main photos come first and the rest after you confirm. Each shot costs one credit across all its ad sizes, and a photo marked Not right? is redone free once. The test kit is free on every plan: photos grouped by version, ad names with UTM tags, Shopify product drafts, AI disclosure wording and pre-order wording with your ship date. When you mark a winner, Pro and above adds a one-page spec sheet for your factory.
Every planSee your next style on a model — before you make it.
Upload one photo of your bestseller. Test new colours, patterns and details on the same model, in every ad size, before you order a sample.
Sources
- [1]Shopify Blog · Product Validation: 9 Proven Strategies (updated 24 June 2025)www.shopify.com/blog/validate-product-ideas ↗ · checked Oct 22, 2026
- [2]Shopify Blog · Why (and How) You Should Test Before You Invest in a T-Shirt Design (3 June 2018)www.shopify.com/blog/sugoi-shirts-testing-process ↗ · checked Oct 22, 2026
- [3]Meta for Business · Ad Measurement: A/B Testing Ads on Facebook & Instagramwww.facebook.com/business/measurement/ab-testing ↗ · checked Oct 22, 2026
- [4]US Federal Trade Commission · Business Guide to the FTC’s Mail, Internet, or Telephone Order Merchandise Rule