Why should an ad that performs well be realized by copying an ad set instead of copying it to another CBO and increasing the budget?

For example, when we launch Facebook ads, the strategy we use is:

CBO 1-10 ADSET (Five ads perform well in it.)

CBO 2-10 ADSET (Keep five good ads from CBO 1, but copy each one again.)

Why should the launching strategy above copy ads instead of directly doubling the budget of CBO 2? Are you worried that the audience will see the advertisement twice?

Before answering this question, you need to know how Facebook advertising finds a matching audience?

This part must not be understood by many people, let’s popularize it with everyone here.

P.S.: About what is CBO, the principle of CBO, the benefits of CBO, and the points to consider for CBO are not discussed here. You can learn them by yourself.

How do Facebook ads find a matching target audience?

Facebook’s algorithm at the ad group level is a bit like this: when launching ads, the ad audience in the ad series will be assigned to an ad. These audiences are tested in the early stage. In the Facebook system, the testing stage is a dynamic matching process (there are matching standards within Facebook). According to this standard, we can determine which audiences have responded to your advertisement.

As we know, Facebook ad’s algorithm uses the Predictive Algorithm.

In short, the machine learning algorithm predicts the effect of new advertising by “learning” the feedback (historical data) obtained from the advertising launched. There are two categories of machine learning algorithms: regression algorithm and classification algorithm.
The result of the regression algorithm is some continuous values, such as a straight line in the quadratic equation and an x value of any abscissa. A corresponding y value can be found.

The output result of the classification algorithm is not continuous, but interval after interval.
For example, when you ask “Will this user click to purchase my product after seeing the advertisement”?
Through analysis, the classification algorithm will tell you “yes” or “no”. But the regression algorithm will tell you that “There’s only 68.59% possibility that the audience will buy, and 31.41% possibility that the audience will not buy”.
No matter which algorithm is used, in the field of advertising, the core of machine learning is to predict the behavior of the audience by analyzing their characteristics (demographics).

For instance:

There’s an ad A in the ad group. The only interest set by the audience of ad A is phone cases.
On the first day, Facebook’s ads may be aimed at this audience’s interest, pushing it to some users who like phone cases and dogs, and observing the performance.
On the second day, Facebook may push the advertisement to another group of different audiences. At this time, these people are not users who like phone cases and dogs, but most likely those who like phone cases and cats.

Facebook will start ad optimization according to the different performances of these two groups of audiences. This optimization is combined with your website’s advertising pixel data to analyze which part of the audience is most vulnerable to your advertisement.
This is the optimization test of Facebook ads. Of course, the internal algorithms are certainly not simple to test one by one. Facebook has a clear grasp of the characteristics of users and will test them in multiple dimensions at the same time.
This is like Facebook will send several advance troops at the same time to spy on the front (according to the audience who may like your product and website pixels) to find the most suitable audience. Several advance teams came back to report to see which performed best and then increase the budget for the force.

Well, after explaining how Facebook ads find a matching audience, let’s solve the problem above. Why increase the budget in the form of copying ads? Why is it implemented in the form of copying an ad set instead of copying them to another CBO and increasing the budget?

Copying ads is mainly because Facebook ads need a learning phase. Copying ads will not interrupt the learning time of the original ads, but adjusting the budget interrupts the learning time of the ads.

For example: For a profitable advertisement, the daily budget is $250. Now the advertisement performs well. You adjust the budget to $500. At this time, the profitable advertisement needs to be studied again, and the original good effect may be interrupted.

Please note that to copy ads, you will need to copy the ad series together instead of directly copying an ad in the original advertising series.

Click on the link to view Chinese version：
https://veryfb.com/d/23-facebook