At the end of October I ran some automated bidding tests on a large Google Ads account. This account is big, so there’s plenty of data to work with, and I specifically tested the Target CPA bidding strategy on some campaigns that had not been performing well inside of this account.
The results were mixed, most of the campaigns had done better with my manual bidding, but a couple of the campaigns had stand out performance with Target CPA bidding, and the conversion rates and conversions are up and the cost per conversions are down. I’m going to leave Target CPA bidding in place for now on the campaigns where it’s working great.
But what’s more interesting than the results of this test, are the two new uses for automated bidding that I discovered while running this test.
A Second Set Of Eyes
The first thing I realized is that target CPA bidding can be a great way to shake up a campaign and get a different look at it.
When you run a big campaign for a long time, there can be certain corners of a campaign that you neglect and forget about. For example, if you’ve got a 20 ad group campaign based on themes, and there’s an ad group in there that is not a big priority for the advertiser, and say it had a bad couple of months with its cost per conversion, you might have bid down the ad group to get a better cost per conversion. But you might have bid the ad group down too low and then started getting a very small amount of traffic to that ad group, and since it’s a low priority, you might have just let the low bids sit there for months and deprived the ad group of its potential.
But then all of a sudden, you throw the entire campaign into Target CPA and the AI will for sure see things that you missed. Maybe you won’t like what it does with those things, or maybe you will, but it’ll see things differently, that’s for sure.
So the AI comes into the campaign and might say “hey, this ad group actually has lots of potential for the cost per conversion that you want, it’s just that you under bid it and neglected it and missed how good it can perform if you just put a medium level bid on it, so I’ll do that.” And them boom, all of a sudden the ad group is performing again and bringing in conversions.
And that’s the first new way I’m thinking about how automated bidding, specifically Target CPA bidding, can come into a campaign and help you. It can provide a fresh look at things. This won’t matter much on a 5 ad group campaign. But for big campaigns with 20 to 50 ad groups, there’s likely some ad groups or keywords that the AI can look at in a new way and help show you some possibilities that you’ve been missing.
Learn From AI Then Control With Manual
The other new outlook that I’m approaching automated bidding with is that it can help show me what’s working in a campaign, and then if I want to, I can flip back to manual bids and continue doing what the AI found, but I can exploit it differently and with more control with my manual bidding.
Sometimes the AI that runs automated bidding shows me something new and finds ad groups, ads, and keywords that can perform better than I thought they could with different kinds of bids and/or by serving the ads to different kinds of people (demographics/audiences), at different times (ad schedule), and in different places (networks and devices). Once I learn something cool from the AI and find something that’s working, I can then take that win, move back to manual bids, and further exploit and control that win, exactly how I want to, with my manual bids.
For example, say I’m getting a $50 cost per conversion. Then the AI comes into the campaign with automated bidding and gets my cost per conversion down to $25 over the course of two months. Then I review the data and I find that the big performance gains came from focusing on desktop ads and focusing on people age 35 and older.
Once I understand what the AI found, I can then take that knowledge with me when I go back to manual bidding, and I can have the insight of the AI and use what it found but also keep the control that comes with manual bidding. I can focus the campaign on desktop and people older than age 35, but my manual bids can quickly dial things up and down as the advertiser’s business goals, situation, and desired aggression levels change from week to week and month to month.
Experiments are the best way to use this back and forth learn from AI, and then control and exploit things back in manual, strategy. With experiments you can easily test out strategies without messing up your existing manual bids, monitor the new data in the experiment, and then shut the experiment down when you’re done with it.
As Google’s Smart Bidding and artificial intelligence get stronger and stronger, I’ll be using them for direct performance gains, but also as a provider of information that I can learn from, get insights from, and further exploit with my unique experience and knowledge about my clients’ businesses, markets, and advertising goals.