Most brands I speak to think their creative problem is a production problem. They want more shoots, a bigger agency, a proper studio day. Usually the real problem is that nobody has a system. Ads go live in random batches, nobody writes down what each one was meant to test, and when performance dips the answer is "we need new creative" with no idea what that creative should actually say. A good creative testing framework for paid social fixes that, and it runs perfectly well on a phone, a freelance designer and a few focused hours a week.
At Money Clothing, paid social ran alongside PPC and affiliate, with spend at £10-20k a month normally and six figures in peak sales months. At that level you can't wait for a studio shoot every time an ad tires, and you can't rely on one hero ad to carry a month. You need a steady supply of new creative and a way of knowing which parts of it are working. That's what this system is for.
Split every ad into angle, hook and format
The first change is how you think about an ad. Stop treating it as one thing. Every ad is three decisions stacked on top of each other.
- Angle. The reason someone should buy. Price, quality, gifting, status, solving a specific annoyance, social proof.
- Hook. The first two or three seconds of a video, or the headline and image on a static. Its only job is to stop the scroll.
- Format. Static image, carousel, UGC-style video, founder to camera, product demo, before and after.
When an ad wins, you want to know which of the three did the work. When it loses, same question. If you change all three at once between ads, you learn nothing you can reuse.
My rule of thumb is to test angles first, because they're the biggest lever. A great hook on the wrong angle still loses. Once you've found an angle that converts, run new hooks against it. Formats come last, mostly to stretch a winner further and fight fatigue.
Keep a simple spreadsheet. One row per ad, with columns for angle, hook, format, launch date, spend and result. It feels like admin. It's the single most useful document your paid social will ever produce.
Borrow winning angles before you invent new ones
Small teams waste a lot of energy trying to be original. You don't need original. You need proven, then made your own.
There are plenty of places to find angles that already work:
- Your own history. Go back through the last year of ads and pull out the top performers. Write down the angle in one sentence. You'll often find two or three themes doing most of the work.
- Reviews and customer emails. Customers tell you exactly why they bought, in their own words. Those words make better hooks than anything written in a meeting.
- Competitors' ads. The Meta Ad Library shows what other brands are running. If a competitor has kept the same angle live for months, it's very likely paying for itself.
- Other channels. Winning email subject lines, high-converting search terms and your best landing page headlines are all angles that have already been tested on real customers.
The important bit is borrowing the angle, not the execution. Copying someone's ad frame by frame is lazy and it rarely works, because your brand, product and audience are different. Take the underlying reason to buy and express it in your voice, with your product and your customers.
Refresh creative weekly, on a fixed rhythm
Creative fatigue is the quiet killer of paid social. The ad that crushed it last month slowly costs more and delivers less, and because the decline is gradual, nobody notices until the monthly numbers look bad.
The fix is boring and it works. Ship new creative every single week, on the same days.
A rhythm I like for a small team:
- Monday: review last week's tests and update the spreadsheet.
- Tuesday: write briefs for this week's batch, based on what you learned.
- Wednesday and Thursday: make the assets.
- Friday: launch, so tests have the weekend to gather data.
As a rough guide, a brand spending low five figures a month can comfortably test three to five new ads a week. Spending less, test fewer, because each ad needs enough budget to tell you something. Ten ads on a tiny budget means ten inconclusive results.
For budget, I use a loose 70/20/10 split. Around 70% goes to proven winners, 20% to iterations on those winners (new hooks, new formats, same angle) and 10% to completely new angles. The new angles are where your next big winner comes from, but they're also where most of the misses live, so cap the exposure.
Structure tests so you can actually read them
A testing system only works if the results are clean. A few rules keep them that way.
Keep testing separate from scaling. Run new ads in a dedicated testing campaign or ad set, then move winners into your main campaign once they've proven themselves. Mixing untested ads into a campaign full of established winners means the algorithm will starve them of spend, and you'll never know if they were any good.
Keep targeting broad and consistent across tests. If one ad goes to a lookalike and another goes to interests, you're testing audiences as much as creative.
Change one thing at a time. If you're testing hooks, keep the angle and the body of the ad the same. If you're testing angles, keep the format the same.
Give each ad enough spend before you judge it. A practical rule of thumb is to let an ad spend two to three times your target cost per acquisition before calling it. Less than that and you're reading noise.
Don't obsess over statistical significance at small budgets. You won't get it. Look for clear gaps instead. An ad that's converting at half the cost of the others is telling you something. One that's 5% better is probably a coin toss.
Make the assets without a studio
This is where most brands assume they need serious money. They don't.
A recent phone, natural light by a window and a clean background will produce video that performs on paid social, often better than polished studio work, because it looks like the content people actually scroll past on their feeds. Staff, founders and happy customers all make good on-camera talent.
UGC creators are worth paying for once you know your angles. Brief them tightly: the angle, the hook you want them to open with, the points to cover and the length. Ask for two or three hook variations on the same video, which gives you hook tests almost for free.
For statics, build a handful of templates with a freelance designer and reuse them. Swap the headline, the product shot and the proof point. One good template can carry dozens of tests.
AI tools help with the grunt work. I build my own AI tools for content and research, and this is a natural fit for them: pulling recurring themes out of hundreds of reviews, drafting hook variations against an angle, writing first-pass scripts and resizing assets for different placements. Treat the output as a starting point. A human with taste still decides what goes live.
What to measure, and when to kill an ad
Split your metrics into leading and lagging.
Leading metrics tell you early whether the creative is doing its job. For video, look at the three-second view rate (is the hook stopping people?) and hold rate (are they watching past the hook?). For everything, look at click-through rate. These settle quickly, so they're useful for spotting a dud hook within a few days.
Lagging metrics tell you whether it makes money: cost per acquisition, ROAS and, ideally, contribution after product and fulfilment costs. These are what decide whether an ad graduates to scaling.
Set kill rules in advance so you're not making emotional calls. For example: if an ad has spent two to three times target CPA with no purchase, it's off. If the hook rate is well below your account average after a few days, it's off. When a winner starts to fatigue, you'll see frequency climb, click-through drop and CPA creep up over a week or two. That's your cue to run new hooks or formats on the same angle, not to abandon the angle altogether.
Whatever happens, write the learning down. "Gifting angle beat quality angle for this product" is worth more than any single ad.
FAQ
How many ads should I test each week?
For most brands spending low five figures a month, three to five new ads a week is plenty. The limit is budget, not ideas, because each ad needs enough spend to give a readable result. If your budget is smaller, test fewer ads and give each one a fair run.
How long should a creative test run?
Run it until it has spent around two to three times your target CPA, or for about a week, whichever gives you a clearer picture. Hook metrics settle within a few days, but purchase data needs longer. Avoid switching ads off after a day because of one bad morning.
Do I need UGC creators to make this work?
No. Staff, founders and customers filmed on a phone can carry a lot of testing on their own. UGC creators become worth the money once you've found angles that convert and want more volume and variety around them.
You don't need a studio to run creative testing well. You need a clear way of breaking ads down, a habit of borrowing what already works, a weekly rhythm and the discipline to write down what you learn. Put those in place and the production side gets much easier. I'm currently open to Head of Marketing and Head of Growth roles, especially with DTC and consumer brands, and I take on consulting work too. If you want help setting up a system like this, get in touch via the contact link on tomosborne.me.