In the era of AI-driven advertising, many media buyers still rely on outdated, manual optimization tactics. A common but flawed strategy involves launching a Campaign Budget Optimization (CBO) campaign with 5 to 6 creatives, running it for a week, pausing the underperforming ads, and then copy-pasting the "winners" into a new ad set to test against new creatives.

In the modern Meta Ads ecosystem, this manual approach is highly counterproductive. Here is a breakdown of why this strategy fails during the cold start phase and how you should structure your testing for predictable scaling.

The CBO Trap: Winner-Take-All and Budget Starvation

When you run a CBO campaign with 5 to 6 creatives on a limited budget (for example, $50 a day), the algorithm does not distribute exposure evenly. Meta's system prioritizes rapid delivery. Within the first few hours, it will allocate the vast majority of the budget to the creative that secures an early lead in Click-Through Rate (CTR).

The remaining creatives are starved of budget, often receiving only a few cents of exposure before being sidelined. If you rely on this method, your "winning" creative isn't necessarily the best performer—it is simply the one the algorithm chose during a period of extreme data scarcity.

The Testing Rule: If you are working with a limited budget, avoid CBO for testing. Instead, use Ad Set Budget Optimization (ABO) or Dynamic Creative (DCT) to guarantee equal exposure for each creative asset.

The Copy-Paste Fallacy: Breaking the Algorithm's Learning Loop

A frequent mistake among novice media buyers is assuming that a creative's performance history or "weight" carries over when copied from one ad set to another.

Meta’s algorithm builds its optimization models based on event postbacks. When you create a new ad set—even if you use the exact same creative asset—the algorithm treats it as a completely new entity with zero history. Every time you copy and paste a creative into a new ad set, you reset the machine learning process.

Instead of constantly duplicating assets, you should:

  • Keep your primary, high-performing ad sets active to maintain learning stability.
  • Use the Post ID of your winning creative to aggregate social proof.
  • Accumulate likes, comments, and shares on a single Post ID to build trust and improve conversion rates, especially for high-ticket offers.

Don't Treat Target ROAS (tROAS) as a Wish List

Another common pitfall is enabling Target ROAS (tROAS) on new accounts with low daily budgets in an attempt to guarantee profitability.

Using tROAS requires a substantial volume of historical conversion data. If you apply tROAS to a new account with a $50 daily budget, the system will struggle to deliver your ads. The algorithm recognizes that it cannot find high-value buyers matching your target return with such limited data, resulting in under-delivery or a complete halt in ad spend.

To scale successfully, adopt a hybrid attribution mindset. Do not rely solely on Meta's real-time ROAS reporting, which is often incomplete due to privacy limitations. Instead, focus on keeping your front-end Cost Per Lead (CPL) within a profitable threshold while optimizing your back-end email marketing and sales funnels to drive final conversions.

Performance marketing is a precise data experiment, not a lottery. By avoiding these common cold-start traps and building stable campaign structures, you can leverage Meta's algorithm to achieve predictable, scalable growth.