One of the most common sources of misunderstanding among meta-advertisers is the dynamic allocation of assets around budget allocation to various assets. meta calls this misunderstanding “Crash effect.”
The danger of the collapse effect is that it convinces advertisers that their budget is being abandoned. As a result, they may take manual steps to cover meta-automation, which often leads to performance efficiency.
This phenomenon is very common. In this post, I will help you get a better idea of what is really going on, provide specific examples of the decomposition effect, and explain how to deal with this situation.
Let’s start…
Explain the effect explained
In the early stages of the campaign, Meta’s ad delivery system entered the so-called learning stage. This is when it tests different assets to determine which assets are most likely to help you achieve your performance goals. The results you see earlier are based on small sample sizes and may not be predictable how the situation will behave over time.
At first it seems that one asset obviously outweighs the others. However, with the advent of more and more data, Meta’s systems can detect trends and changes in performance. It then starts allocating more budget to the assets it predicts, even if the asset does not have the lowest cost per acquisition (CPA) at the beginning.
Since you tend to get up more money on any given asset, an asset that looks best (but has a smaller budget) in the early stages may end up with a lower average CPA than a more budgeted asset. This seems to make Meta prefer the “worse” performers, when in fact, the system is trying to maximize your overall score throughout the campaign.
Meta said that the misunderstanding of budget allocation and the presumption of improper budget are inappropriate. Fault effect. Let’s discuss some examples.
Example: Location
META uses budget allocations between placements as an example of the failure effects in its official documentation. To avoid unnecessary complexity, it provides an oversimplified example that is clear only with two locations.
- Budget: $500
- Location: Facebook Stories and Instagram Stories
In the early stages, Facebook stories were significantly better than Instagram stories. However, there was a turning point on Day 4, and Instagram stories began to outperform Facebook stories. This is a visualization of the trend.
Meta starts shifting more budgets to Instagram stories. As spending increases, costs naturally increase. In the $500 total budget, $450 will be used for Instagram stories, while only $50 in Facebook stories. When looking at only the total average CPA, it may seem that more money will be spent on lower performance positions. This is Fault effect.
Since these are averages, they missed the trend. When the day breaks down, it’s clear that Instagram stories are the best performers over time.
Although this example makes the trend obvious, you won’t always point out such obvious changes in CPA. The budget allocation of the system can also be driven by the projected CPA as expenditure increases, not just historical results.
Consider another example: two locations with different inventory potential. Suppose you use both Facebook feed and Threads feed locations. Facebook has about 3 billion active users, while threads are only 350 million.
From the thread feed, you may get a lower CPA than the Facebook feed. However, Meta may spend more on Facebook feeds due to the larger stock available. If you force all delivery to thread feeds, your costs may rise as the limited audience gets exhausted.
The misunderstanding that Meta misallocates more budget to Facebook feeds compared to threaded feeds is Fault effect.
Example: Advertising set
Similar phenomena may occur when using multiple ad sets with advantageous campaign budgets. Suppose you are using two ad sets, each with a different targeting method:
You allocate a $100 daily budget and Meta will be dynamically distributed between two ad sets, resulting in the best results for you. But a few days later, even though its CPA was twice as high, META had allocated 80% of its budget to AD Set 1.
- Advertising set 1: $4.00 CPA (80% of budget allocation)
- Advertising set 2: $2.00 CPA (20% of budget allocation)
Did Meta mistakenly allocate more budgets to underperforming ad sets? This misunderstanding is Fault effect.
Meta’s ad delivery system recognizes that the potential audience ranges in these two ad clusters are very different. Remarketing ads have a limited audience, which means you will quickly reach a decrease in returns and the cost of increasing as you try to scale. Even if the cost per conversion is higher, the limiting balance between algorithms and larger audiences in a wide range of target ad clusters.
Ultimately, the algorithm is driven by the overall goal of getting the most conversions throughout your campaign. It may be predicted that increasing the budget for the remarketing ad set will actually result in higher costs and fewer total conversions.
Example: Advertising
This is a common situation. You have two ads in the same ad set. Early in the testing, the budget was evenly distributed between AD#1 and AD#2, and AD#1 reached a lower CPA. But as the days went by, META’s system found that AD#1’s CPA began to surpass AD#2’s CPA, so the budget was larger and dedicated to AD#2.
As more budgets are allocated to AD#2, its costs rise, resulting in higher CPAs. By the end, AD#2 had consumed 80% of its budget, but its overall CPA now exceeds AD#1’s budget.
Advertisers turned off AD#2 when they saw that Meta seemed to misallocate budgets to the “error” ad. This is a classic example Fault effect.
Costs may increase when advertisers close the second ad and serve all budgets to AD #1. The end result will be higher than the overall CPA when using two ads with META automatic budget allocation.
Meta’s goal is to provide you with the largest aggregated results in all ads in the ad set, as defined by your performance goals. The system balances trends and forecasts the cost of all ads to achieve this.
Recommended method
A common topic in my suggestion is reflected here: Prioritize handheld measures over micromanagement. The average CPA at the surface level can be exceeded when evaluating why the budget is allocated as is.
Focus on results totalinstead of isolating a single asset. For example, use:
1. Multiple ad sets with advantageous campaign budgets: Prioritize aggregate activity performance over the performance of a single ad set.
2. Advantages + Position: Prioritize the performance of aggregated ad sets over the performance of a single location.
3. Multiple ads in the ad cluster: Prioritize the performance of aggregated ad sets over the performance of individual ads.
This does not mean that the algorithm is perfect, but continuous micromanagement of budget allocations is unlikely to improve performance. In most cases, it actually makes delivery less efficient.
now you
Do you have any other examples of decomposition effects?
Let me know in the comments below!
