Testing Framework: Find Winners & Kill Losers

Most people run one ad and call the whole channel a failure the ones who build profitable ad accounts run dozens of variations systematically and iterate toward a profitable CPL

An ad account is not a campaign. It is a process of iteration toward a profitable cost per lead.

The 60/30/10 Budget Rule

Once you have data from your first tests allocate your budget in thirds.

60% behind proven winners which are ads with a profitable CPL that have been running for at least 2 weeks. Scale these by increasing budget or duplicating the ad set.

30% behind close variations of winners using the same hook with different creative or the same creative with a different CTA. Small changes to proven performers.

10% on genuinely new swings covering a completely different hook angle, a different format, or a different audience. Most of these will not win. One occasionally does and that one becomes the next winner.

This allocation keeps revenue stable from the 60% while continuously improving from the 30% and discovering new winners from the 10%.

At the start when you have no winners yet your entire budget is in the 10% category because you are doing pure testing. That is correct. You are buying information.

What to Test and in What Order

Do not test everything simultaneously. The order matters.

Round 1: Test the hook and angle which is the most important variable. Run 3 to 4 different hooks and angles with the same creative and CTA. Which problem angle resonates? Which hook type gets the best CTR?

Round 2: Test the creative given a winning angle. Keep the same hook and copy and test different images covering before and after, screenshot, text on background, and talking head. Which visual stops the scroll?

Round 3: Test the CTA given a winning hook and creative. Keep everything the same and test different CTA text and buttons. "Book a free call" versus "Get your free quote" versus "See how it works."

Round 4: Test the landing page given a profitable ad. Send traffic to two versions of the landing page using a different headline, different form length, or different proof elements.

Each round narrows down to a winner. Each winner becomes the control for the next round.

When to Kill a Loser

Kill an ad when it has spent £30 to £50 and the CTR is below 0.5% meaning no one is clicking and the hook is wrong, or the CPL is 3x or more above your ceiling meaning people are clicking but not converting and the landing page or offer is wrong.

Do not kill before £30 to £50 spend because there is too little data and normal variation will look like failure on day 2.

Do not keep running past £100 if CPL is consistently bad. The data is telling you something is fundamentally wrong with the hook, the creative, or the audience.

When to Scale a Winner

Scale an ad when it has at minimum £50 to £100 of spend with a profitable CPL, at least 5 conversions which means leads and not just clicks, and has run for at least 7 days to account for day-of-week variation.

How to scale: increase budget by 20 to 30% per week and not more because doubling the budget often resets the algorithm and tanks performance. Duplicate the winning ad set and run the same ad in a new ad set at a slightly higher budget letting both run in parallel. And create variations of the winning creative using the same angle with slightly different execution.

Using Claude for Testing Analysis

After your first 7 to 10 days of data run this prompt:

Using @ADS.md — help me analyse my first Meta ad campaign results and

decide what to do next.

Campaign: [NAME]

Budget: £[X] per day

Days running: [X]

Results so far:

- Ad A ([HOOK TYPE]): Impressions [X] · CTR [%] · CPL £[X] · Leads [X]
- Ad B ([HOOK TYPE]): Impressions [X] · CTR [%] · CPL £[X] · Leads [X]
- Ad C ([HOOK TYPE]): Impressions [X] · CTR [%] · CPL £[X] · Leads [X]

My target CPL ceiling: £[X] (10% of average project value from ADS.md)

For each ad tell me:

1. Keep, Pause, or Scale — and why
2. What the CTR data is telling me about the hook
3. What the CPL data is telling me about the landing page and offer
4. The single most important thing to test next
5. Draft 3 new hook variations based on what the winning angle suggests

Keep recommendations specific. Do not tell me to "keep testing" without

a specific next action.

The Testing Mindset

Test hypotheses and not ads. Every ad test should be answering a question such as "Does the problem angle perform better than the outcome angle?" and not just "which ad is better?" Name your tests after the hypothesis.

Test one variable at a time. If you change the hook and the image simultaneously you do not know which changed performance. Discipline here saves weeks of confusion.

Kill fast and scale slow. Most ads fail and that is normal. Kill the losers after £30 to £50. Scale the winners gradually. The winners last weeks and the losers cost you £30 to learn from.

Losers teach more than winners. A hook that gets 0.4% CTR tells you that problem angle does not land with this audience. Write that down because it changes how you write the next batch.

Building Your Testing Log

Add to ADS.md after every test:

## Testing Log

Test 1 — Hook angle (Week 1 July 2026)

Hypothesis: Problem angle outperforms outcome angle

• Ad A (problem): CTR 2.1% · CPL £28 → WINNER
• Ad B (outcome): CTR 0.9% · CPL £67 → PAUSED
• Ad C (question): CTR 1.4% · CPL £41 → WATCHING

Conclusion: Problem angle wins with this audience. Test creative next.

Test 2 — Creative format (Week 2 July 2026)

Hypothesis: Before/after screenshot outperforms text-on-background

Winner: [TBD after this week's data]

Your Action Right Now

After your campaign has been running 7 days, pull the data and run the Claude analysis prompt above. Make the decisions it recommends. Do not look at the data before 7 days because you will make emotional decisions based on noise.