On Meta, TikTok and increasingly Google, the ad platforms now handle most targeting automatically. That leaves creative, the images, videos and words in your ads, as the biggest lever you still control. Brands that test creative systematically tend to beat brands with bigger budgets that don't.
Here's a simple system to decide which ads to cut, which to fix and which to scale.
The metrics that matter
| Metric | What it tells you | Where problems show up |
|---|---|---|
| Hook rate (3-second views ÷ impressions) | Does the first second stop the scroll? | Video opening |
| Hold rate (15-second or ThruPlay ÷ 3-second views) | Does the story keep attention? | Middle of the video |
| Click-through rate (CTR) | Does the ad make people want more? | Offer, message, call to action |
| Landing page conversion rate | Does the page deliver on the ad's promise? | Website |
| Cost per purchase / lead | Is it profitable? | Everything combined |
| Return on ad spend (ROAS) | Revenue per pound spent | Everything combined |
The first three diagnose creative. The last two decide what to do.
Set your thresholds first
Before testing, write down:
- Target cost per acquisition (CPA): the most you can pay for a customer and still profit. Our break-even analysis guide helps you calculate it.
- Minimum spend per ad before judging: often two to three times your target CPA.
- Minimum time: at least three days, ideally a week, to smooth out daily swings.
The decision table
Once an ad has had its minimum spend and time:
| Result | Decision |
|---|---|
| CPA well below target, steady for several days | Scale: increase budget gradually, make variations |
| CPA around target | Keep and watch; test new hooks to improve it |
| Good hook and CTR, poor conversions | Fix the landing page or offer, not the ad |
| Weak hook rate | Fix the opening: new first second, new thumbnail |
| Good hook, weak hold | Fix the middle: tighten the story, show the product sooner |
| Spent 3× target CPA with no conversions | Cut |
| Was profitable, now rising CPA and frequency | Refresh: creative fatigue, make new versions |
What to test (in order of impact)
- Concept: the core idea. A problem/solution demo, a customer testimonial, a founder story, a comparison, an unboxing. Concepts make the biggest difference.
- Hook: the first one to three seconds. Same video, different opening.
- Format: short video, static image, carousel, user-generated style.
- Offer and message: price, bundle, free shipping, guarantee.
- Details: colours, captions, music, call-to-action wording.
Test big things first. Changing a button colour rarely matters when the concept doesn't work.
A weekly testing routine
Monday: review last week's ads against the decision table. Cut, fix or scale.
Tuesday: brief three to five new ads, at least one new concept and the rest variations of your current winners.
Wednesday–Thursday: produce and launch.
Friday: quick check for anything badly broken (rejected ads, landing page errors). Don't judge results yet.
Scaling without breaking things
- Increase budgets by around 20–30% at a time, every few days, rather than doubling overnight.
- Make variations of winners: same concept, new hooks, new people, new formats.
- Watch frequency. When the same people see an ad too often, results fall.
- Duplicate winning ads into broader campaigns rather than editing the original heavily.
Where creative ideas come from
- Customer reviews and support emails: the exact words customers use make great hooks.
- Comments on your ads: objections tell you what the next ad should answer.
- Your best-performing organic posts.
- Competitor ad libraries (Meta and TikTok both publish them): look for angles, never copy.
- Your sales team: what do they say that works?
Common mistakes
- Judging after one day. Early results are noisy.
- Testing too many ads with too little budget. Nothing gets enough data.
- Cutting ads that drive good clicks when the real problem is the landing page.
- Never refreshing winners until they collapse.
- Trusting platform-reported ROAS alone. Check real sales in your shop or bookkeeping. Read server-side tracking if numbers don't match.
Let software do the reading
Going through exports by hand every week is slow. Zelvorm Ad Doctor reads your Meta, TikTok and Google Ads exports and tells you exactly which ads to cut, fix and scale, using the same kind of thresholds described here. It's free.
For Google-specific automation, see Google Ads AI Max and Performance Max in 2026.
The short version
Creative is the biggest lever left in automated ad platforms. Set a target CPA and minimum spend before judging, then use hook rate, hold rate and CTR to diagnose and cost per acquisition to decide. Cut ads that spend three times your target with no results, fix the specific weak point of promising ads, scale winners gradually and refresh before fatigue sets in.
Need ads that perform? Our digital marketing team and content creation team plan, produce and test creative together.




