Performance

A simple creative testing framework for Meta and TikTok ads

On Meta and TikTok, targeting is increasingly automated. The platforms decide who sees your ads — which means the creative itself has become your most important targeting tool. The brands that win are the ones that test creative consistently. Here is a framework simple enough to run every week.

Test in the right order

Not all tests are equal. Changing a button colour rarely matters; changing the core message often doubles results. Test from the biggest lever to the smallest:

  1. Concept / angle — the core reason to buy: price, speed, status, problem solved, social proof.
  2. Format — UGC-style video, founder talking to camera, product demo, static image, carousel.
  3. Hook — the first two seconds or the headline. The same video with three different openings can perform completely differently.
  4. Details — copy length, call to action, music, thumbnail.

Structure your account for testing

Keep it simple. Separate your proven ads from your experiments so tests don't steal budget from what already works:

  • Scaling campaign: your best-performing ads, with most of the budget (around 70–80%).
  • Testing campaign: new concepts, with a fixed share of budget (around 20–30%). One ad set per concept, three to five variations inside it.

On TikTok, also test Spark Ads using creator or organic posts — native-looking content usually outperforms polished ads on the platform.

How much to spend per test

A useful rule: give each concept enough budget to reach at least 20–30 conversions, or roughly two to three times your target cost per acquisition before judging it. If that is too expensive, optimise for an earlier signal — such as add-to-cart or lead — while volumes are low.

When to call a winner

Look at the metrics in order, from attention to action:

MetricWhat it tells you
Hook rate (3-second views ÷ impressions)Does the opening stop the scroll?
Hold rate / average watch timeIs the story interesting enough to keep watching?
Click-through rateDoes the ad create enough interest to act?
Cost per acquisition & ROASDoes it actually make money? The only metric that decides.

The early metrics explain why an ad works or fails; the final ones decide whether it stays. An ad with a great hook rate but poor sales may need a better offer or landing page, not a new video.

Scaling what works

  • Move winning ads into the scaling campaign with their post IDs, so you keep the social proof (likes and comments).
  • Increase budgets gradually — around 20–30% every few days — rather than doubling overnight.
  • Make "iterations" of winners: same concept, new hooks, new creators, new formats. Winning concepts usually have many more winning ads inside them.
  • Expect fatigue. Even the best ads decline, which is why testing never stops.

A weekly testing rhythm

  1. Monday: review results and pick next week's concepts.
  2. Tuesday: brief and produce new creative — this is where an AI-assisted content studio saves days.
  3. Wednesday: launch three to five new tests.
  4. Following week: judge, promote winners, cut losers, repeat.

Teams that run this loop every week build a library of proven angles over time — and that library becomes a real competitive advantage.

Want a second opinion on your ads? Our free ads audit reviews your account structure, tracking and creative, and shows the three changes we would make first. Request an audit.

Ready for
lift-off?

Tell us where you want to grow. We reply within one business day with first thoughts and next steps.

[email protected] Working with clients worldwide Reply within 1 business day
  1. We read your message and reply with first thoughts — no automated sales sequence.
  2. A free 20-minute call to understand your goals, markets and timing.
  3. A clear proposal with the scope, deliverables and timeline.
Message received.

Thanks — we'll get back to you within one business day. If it's urgent, email [email protected].

What do you need?

By sending this form you agree to our Privacy Policy. We never share your details.