Entry #86: How to use Facebook’s reporting tool to pivot your data and learn from your previous spend.

This is the 86th post in a series. The first post can be found here.

Progress Report

I spent a good amount of time today working on a new Agency Overdrive lecture that will focus on profitable client billing strategies.

I’ve sent over 500 proposals in the past few years, and that’s a topic I have a lot to say about. I’m excited about making that a series of lectures that I think students will benefit a lot from.

My new Zoom webinar campaign got off to a good start this morning, showing an average CPA of about $4.

But then I messed up the conversion tracking accidentally (long story), and no other conversions were tracked throughout the rest of the day, which of course, completely throws off the entire algorithm.

I just fixed it, and luckily Facebook will backlog conversions accurately, but the average CPA I wound up with was $16 or so. Not good at all.

The $50 / day conversions campaign did well today, with a $5 CPA.

I also realized more and more today how important it is to analyze the previous $50,000 in spend, and see if I can find any meaningful stories in the data. After all, that’s what I do for clients every day.

Facebook has a very useful reporting tool, where you can pivot the data based on breakdowns of your choosing.

So first, I looked at gender and age:

I also didn’t want to just look at registrations. I wanted to look at sales, which is the only metric that actually matters.

As you can see the 25–34 Male segment performed the best, at a $435 average cost per purchase. In fact, that was the only profitable segment, if you analyze each segment separately.

That segment had a 50% lower CPA than all the other segments combined. Females performed abysmally throughout the age segments. I have nothing against women, of course, but I’m not going to spend money on a demographic if it’s not producing any financial returns (I have a wife for that already).

Just based on the above data alone, I decided I should make a new version of the new campaign, which is what I did, but first I looked at some additional breakdowns.

Here’s a look at device platform:

While mobile apps produced the highest volume of sales, desktop produced the only profitable ROAS. So, that’s something. In my new campaign, I’m now only targeting desktop devices.

But to analyze it further, Facebook shows you the device a user was on at the time of conversion — which is pretty cool.

As you can see, the vast majority of conversions took place on a desktop, which makes sense, being that this is a webinar funnel.

This further solidified my idea to only target desktop computers in this next campaign (which will run tomorrow).

I duplicated the original version of this new campaign, I set it to only target males between 24 and 35, only in the US, and used my lookalike audiences (the same ones I had in today’s campaign).

I also edited the placements to only show on Facebook.

Beyond that, I also narrowed the dynamic creative to the strongest assets, as indicated by the historic data in the account.

I’m only running one description and two headlines (and one CTA text).

This wasn’t guesswork. This was all based on the reporting breakdown analyzing $50,000 in spend.

Dynamic creative is a great tool to use, but you can also use it foolishly, if you’re not learning from the historic data in your account.

Facebook is going to give all your variations at least one impression, and you’re much better off giving Facebook your strongest assets (and fewer of them), then giving them a higher volume of weak assets mixed with the strong ones.

This is something worth remembering.

Facebook’s #1 priority with dynamic creative is to test everything, even if that means risking a higher CPA. They are not primarily driven by a low CPA, even if you’ve told Facebook to optimize for lowest cost.

They will optimize for lowest CPA, eventually, but first they want to test as many variations of your creative as possible. So if you know, from previous spend, that certain texts, headlines, images, etc., don’t convert as well, you shouldn’t keep feeding them into Facebook’s algorithms.

Lastly, use common sense when you analyze your breakdowns.

For example, when analyzing which headline performed the best, it would be a mistake to look at purchase rate or ROAS as your deciding metric. While there may be a statistically significant difference between two headlines, it is very much unlikely that the headline variance caused the purchase variance, which happens hours / days later, based on a lot of other factors.

I believe this is called a lack of face validity in statistics, but I don’t remember.

Either way, you need to look at the metric most likely to be affected by the variables you’re analyzing.

So if you’re analyzing headlines, you should look at CTR (click through rate). Not even registration rate. Registration rate could be influenced by the landing page copy, form fields, device and more.

CTR is the most relevant and linear representation of an effective headline, and that’s the metric you should use to determine the winning headline(s).

On the other hand, when analyzing gender, age or device, you want to ignore the meaningless metrics (cost per registration / registration rate), and look at the metrics that really matter most (ROAS / Cost per sale).

If you haven’t toyed around with Facebook’s reporting tool, I highly recommend you give it a spin.

Enjoy it!

And, keep on truckin’
- Isaac

The next post can be found here.