Testing identifies winners. Scaling turns winners into revenue. Killing losers protects your budget. This module covers the operational mechanics of each: when to scale, how much to increase, when to kill, and how to manage the creative refresh cycle that prevents fatigue.
When to scale a winner
A creative variant earns scaling when it meets three criteria simultaneously over a 48-72 hour period. First, the CPA is at or below your target. Second, the CTR is above your category benchmark (1% for cold traffic on Meta is a reasonable floor). Third, the performance is stable: the daily CPA is not swinging wildly between $5 and $50.
Stability is the most overlooked criterion. A variant that hit a $10 CPA on day 1 and a $40 CPA on day 2 has a $25 average, which might meet your target. But the instability means the result is unreliable. Wait one more day. If day 3 stabilizes near the average, scale. If day 3 spikes again, the variant is too volatile to scale safely.
Do not scale a variant that has fewer than 5 conversions. Five conversions is the bare minimum for any confidence in the result. Ten is better. Twenty is ideal. With fewer than 5 conversions, you are making decisions based on noise rather than signal.
How to scale: the 20% rule
Increase budget by no more than 20% per day. Meta's algorithm needs time to recalibrate delivery when budget changes. A 20% increase ($100 to $120) is small enough that the algorithm adjusts smoothly. A 100% increase ($100 to $200) forces the algorithm to find an entirely new audience segment to fill the additional budget, which often means reaching less qualified prospects at higher cost.
The scaling path for a $100/day starting budget: Day 1: $100. Day 3: $120. Day 5: $144. Day 7: $173. Day 10: $207. In 10 days, you have doubled the budget while maintaining performance stability. In 20 days, you are at $400/day. The patience pays off because each increase confirms that the higher budget still delivers acceptable results.
If performance drops after a budget increase, do not immediately cut the budget. Give it 24 hours. The algorithm adjusts. If performance does not recover within 48 hours, roll back to the previous budget level and wait 3 days before trying again. Sometimes the audience simply cannot absorb more budget at the same efficiency, and that ceiling is useful information.
When to kill a loser
Kill a variant when any of these conditions are met: CPA is 2x your target after 48 hours and 1,000+ impressions. CTR is below 0.5% after 48 hours. The variant has spent $50 with zero conversions. These thresholds are aggressive by design. Keeping underperforming variants running is the single biggest waste of ad budget in DTC marketing.
Do not kill a variant after 24 hours unless it has zero clicks from 1,000+ impressions. One bad day does not make a bad ad. Two bad days usually does. The 48-hour minimum prevents premature kills driven by normal performance variance.
When you kill a variant, document what you learned. "Question hook with pricing comparison did not resonate with cold audience on Meta Feed. CPA $42 vs $18 target after 72 hours." This documentation prevents you from retesting a proven loser in the future.
Creative fatigue and the refresh cycle
Every winning creative has a lifespan. On Meta, the typical lifespan is 2-4 weeks before performance degrades. The signals of creative fatigue are: rising CPM (the platform is charging more to show your ad because engagement is declining), declining CTR (the same people are seeing the ad and no longer clicking), and rising CPA (fewer conversions per dollar spent).
Monitor these three metrics weekly for every active creative. When all three move in the wrong direction for 3 consecutive days, the creative is fatiguing. You have two options: pause and replace (stop the fatiguing ad and launch a new creative in its place), or iterate (generate variants that preserve the winning concept but change the execution, such as a new hook on the same visual or a new CTA on the same copy).
The refresh cadence for most DTC brands: launch 5-10 new creative variants per week, regardless of how well current ads are performing. This ensures you always have fresh creative entering the testing pipeline before current winners fatigue. Mani's daily queue is designed for exactly this cadence: 3-5 fresh variants every morning, reviewed in 15 minutes, exported in one click.
The creative portfolio approach
Do not depend on a single winning creative. Maintain a portfolio of 5-10 active creatives at all times, each serving a different audience segment or funnel stage. If one creative fatigues, the others continue performing. The portfolio approach smooths out the revenue impact of individual creative failures.
The ideal portfolio: 2-3 creatives for cold audience (awareness), 2-3 for warm audience (consideration), and 2-3 for hot audience (retargeting). Each creative uses a different hook pattern, visual style, or messaging angle. The portfolio creates diversity in what your audience sees, which extends the effective lifespan of each individual creative because the same person never sees the same ad twice in a row.
This portfolio approach is the core philosophy behind Mani's generation model. Rather than creating one perfect ad, Mani generates multiple variants across formats and angles. You approve the winners, export them, and let the portfolio work while Mani generates the next batch. The creative pipeline never stops, and your ad account always has fresh material to test.
This concludes the AI Ad Creative Fundamentals course. You now have the framework for creating, testing, scaling, and refreshing ad creative systematically. Apply these principles using Mani at /playground or sign up at maniai.com.