Statistical Noise
Statistical Noise is making sure that the metrics you are comparing and optimizing are significant enough to make decisions on. To ensure we are focusing on scalable metrics, we should have achieved one of the following after two weeks:
● 10,000 impressions.
● 25 link clicks.
This ensures that the ads have been shown to a wide enough audience for Facebook to gauge how well the content & audience target match — and how heavily they will penalize us through CPM, CPC, CPA, and other cost metrics.
Facebook has told us directly that the learning process of ads, creatives, objectives, and audience takes a minimum of two weeks. Once the algorithm has gathered enough data, then it will be able to optimize according to all of the different variables.
Examples where Statistical Noise is most dangerous:
Example 1
We’ve run our campaign for 2 days, it’s picked up 200 impressions, and it’s already closed $250 in sales from $4 in ad spend.
The natural thing for most people to do is ramp up spend to $400 since if we increase spend to $400, our sales will go to $25,000. Right?
Wrong.
Seeing as we are spending $2 a day on this particular campaign, we’ll let it run for the 2-week learning phase — spend $30, and then we can determine what our next optimization steps are and if it continues to generate revenue at a 63x ROAS.
Example 2
Our campaign has been running for 4 days now. It’s only reached 500 people, but Facebook is reporting our CPM as $23.
It seems like our audience is really expensive, so we would think to go in and turn the campaign off before it wastes any more of our dollars.
But when you put this into perspective, we’ve spent just $11.50 over 4 days and we haven’t given this enough time for Facebook to learn.
We are spending $2.88 a day on this campaign, so it’s suggested to let this one run until it picks up enough results to make a more educated decision based on our Standards of Excellence.
Example 3
We’ve spent $10 on a campaign and generated 2 link clicks. We are aiming for $0.50 per link click, and we are currently achieving costs 10x this.
We want to go in and turn the campaign off because it is costing us too much per link click, but we’ve only been running the campaign for 3 days, so we haven’t given it enough time to collect results and properly optimize.