The audience signal lie

The creation of this post is based on multiple account audits and instances where I have read that people have been living under the impression that they could mold a Performance Max campaign to only serve ads to people who have interacted with your business in a certain way and have been trying to run dedicated remarketing Performance Max campaigns/asset groups by just added "People who did not complete checkout" audience signals.

This does not work, or make sense at all when you know how audience signals work and operate.

How audience signals work

Let's begin with some Google documentation on how audience signals are described

"Audience signals allow you to add audience suggestions that help Google AI optimize for your selected goals. Although adding audience signals is optional, using audience signals can help you guide machine learning models on the ideal way to optimize your campaign"

This functionality of audience signals is just as it says - a signal.

It's a way for us to feed Performance Max with our 1st party data and help/guide smart bidding in the right direction and help it reach its bidding target (Target ROAS or CPA) 

Smart Bidding already knows to an extent how the customer demographic might look like and will be able to capture conversions, but an audience signal can help shorten the gap between a new campaign launch and till it pumps out a desent conversion volume

The example below visualizes that process.

So how do i work with audience signals?

Start by looking at which 1st party dataset you are strongest.

If you have a customer list of let's say +3000 individual customers I would prioritize that and use that as an audience signal, but if you don't have that much customer data you can move it down to a list of people who started checkout but did not purchase and use that as audience signal until you have +1000 individual customers (you need +1000 individual customers on a customer list to be able to upload it to Google Ads) 

If you don't have any data at all, you could begin to look into in-market audiences or custom segments, but I would recommend exploring your own 1st party data before digging into those audience types.

If you have more than 5000 customers historically you could take it a step further and segment your customer list into being "customers who purchased for more than €80" and only use customer lists of your very best customers to guide smart bidding in the direction towards your high AOV customers.

If you are present in more markets than 1, you could also use country-specific customer lists, and segment by that way.

It all comes down to how much data you have and your ability to slice it into fitting Performance Max campaigns.

Can you A/B test audience signals? In theory yes, but there is close to zero value nor learnings in split-testing a signal at all. The time you could spend creating audience signal A/B test can be applied better elsewhere

Takeaways

1. Use your audience signals as signals for your Performance Max campaign (duh) 
2. Use 1st Customer Lists as audience signals if you have significant enough data
3. Do not A/B test audience signals (it's a waste of your time) 

Hope this give you some idea on how audience signals work, and has enabled you to have a clear process on how to work with them going forward.