Inside Google Ads podcast: Episode 114 - Learning periods

“I don't want to do that, it's going to trigger the learning period!”

“No, I don't want my campaign to go back into learning!”

“Why is my campaign still stuck in learning?”

My goodness. If there's one thing I've noticed recently that a lot of practitioners are… I don't want to say freaking out about, but definitely worrying about, it's the dreaded learning period. 

Now I have said before, your campaigns are always learning. There's not like some set period, boom, they've learned, now they're done. Just like human beings are, hopefully, always learning. It's not like one day we've woken up and we've learned everything there is to know and we're perfect now.

But I was recently listening to an episode of Ads Decoded, an amazing official Google Ads podcast hosted by Ginny Marvin. And in this episode, Ginny was interviewing Brandon Ervin, Director of Product Management for Search Ads at Google. And he dropped a lot of really interesting nuggets that are reframing the way I think about learning. So I had to share that with you.

In this episode, we're going to bust three common myths I hear about learning periods in Google Ads so that you can feel more confident making the changes you need to make.

I'm your host, Jyll Saskin Gales. I spent six years working for big brands at Google, and now I work for you. 

This is Inside Google Ads, Episode 114: Learning Periods.

1. The first myth I often hear about learning is that people don't want to touch anything in their Google Ads account because they're scared it will trigger learning. 

What Brandon shared on Ads Decoded is that Google's models are semantic-based rather than syntactic-based.

What that means is that Google understands the actual content and context of the building blocks of your campaigns, even if you move them around. What he said is, “We prioritize putting what we would call a semantic feature into the model and not a feature like campaign ID or ad group ID.”

If you move your creatives from one ad group to another, it doesn't have to relearn. And this contradicts what a lot of us think about. “If I add a new keyword or if I change a headline, everything's going to have to relearn.” I always knew that wasn't the case by my own experience. And now the person who designs these products has proven it for us.

So let's run through a quick cheat sheet of what will and won't trigger relearning.

Swapping an asset in your ads or copying and pasting ads between ad groups? No. Brandon confirms the model didn't even know these assets were in different ad groups in the first place. 

Modest budget changes? No. Small tweaks stay within the existing learning that's already happened.

Keyword changes? It depends. If they match your existing theme of the ad group, then the AI probably already has its context. As Brandon says, “Keywords are a means to an end. They aren't the end in and of themselves.” 

Regardless of how you feel about that - I'm sure some people have feelings - that reassures us that changing some keywords around will not have a drastic impact on performance.

Now, the things that can trigger a full relearning are changing a conversion action. This is a substantive change because you really change what the campaign is trying to accomplish.

And a bid strategy change. For example, moving from Max Conversions to Target CPA completely changes the goal.

And what do you know? This is what I've been telling you for many, many years, but great to hear it confirmed by the person who builds these products.

So if you need to make a change or optimization to your campaigns, go ahead, make that change. It generally does not mean your campaign is going to start all over from the newborn phase and have to do everything from scratch.

2. The second myth about learning we're going to bust today is about a magic number. 

I usually say you need 30 in 30, 30 conversions in 30 days for smart bidding to work. A recent Optmyzr study found that most advertisers need 50 conversions in a 30-day period in order to consistently see better performance.

What Brandon from Google shared, however, is that the general guidance they publish is to achieve 15 conversions over a 30-day period - only 15 conversions.

So what's the truth here?

The truth is that there's no one number that magically means if you have 14 conversions, your bidding is dumb, and at 15, it's magically smart. Or at 28 conversions, your bidding is dumb, and at 30, it magically works. Your campaign is consistently learning. 

The way I would think about this is Google's 15 conversions in 30 days is about technical eligibility, a bare minimum. Without at least that, there's not really much smart bidding can do. 

Whereas, on the other end of the spectrum, Optmyzr's 50 in 30 is much more about statistical consistency. If you have at least 50 conversions in 30 days, it will be nearly impossible for smart bidding to lose.

The 30 in 30 that I preach is somewhere in the middle. And that's just based on what I have seen most consistently work across the thousands of accounts I've seen, the kinds that spend $10 a day, a million dollars a day, and everything in between.

Interestingly, in this episode, in a discussion about campaign structure, Brandon shared the exact advice I've been giving a lot of my coaching clients lately: ‘Don't be scared of leveraging portfolio bid strategies and shared budgets, because when you do that, you're essentially lumping your campaigns together without having to do a full restructure.’ This is a great way to test consolidation and see if that data density leads to better results before taking the plunge and actually restructuring your campaigns.

And again, when you do that, you're not going to trigger a whole ‘relearning from zero’ because as Brandon shared, your campaigns and account already understand those different assets, those different search terms, even if you move them between ad groups or campaigns.

3. Myth number three that we're going to bust today is that there is a set learning period. 

When you do make a significant change to your campaigns, you may see “bid strategy learning” as a status, but no matter what, you're going to see that for five days. It's more of a UI thing than actually telling you what's going on under the hood. It's not like on day six or day eight or day 10, some switch flips and your campaign has magically learned.

Brandon confirmed this, sharing that it's more about the conversion cycles and the data density rather than a time period. So if you run the kind of business where people tend to click on your ads and convert the same day, then learning can happen faster. Whereas, if you're like a lot of my lead gen or B2B clients where someone clicks on an ad and you have a conversion lag of maybe seven days or 14 days, or 40 days… it's going to take a lot longer for that campaign to learn because it's not going to know if the conversion happened for days, weeks, or even months.

This, by the way, is exactly why I recommend starting your campaigns, for the most part, on Maximize Conversions bidding. That means from day one, you are feeding your smart bidding algorithm data and telling it to optimize for conversions. Even if it doesn't yet know what a conversion looks like, as soon as those first conversions start rolling in, it's going to learn right away. Rather than starting on something like manual CPC, where all it's trying to do is buy clicks, and you're going to buy a whole bunch of those clicks and then conversions may or may not happen - at some point you're going to switch bid strategy to a smart bidding strategy, which will suddenly trigger new learning because your campaign was never trying to get a conversion, and now all of a sudden it is.

Give your machine that data it needs and let it start learning from day one by starting on a conversion-focused bid strategy. And then once you have enough data density, as Brandon calls it, or just conversions, as I call it, you'll wanna shift to a target-based bid strategy to stabilize your results and enable you to scale.

So to bust that myth, there is no set timeframe for learning. Learning is about the volume of conversion data, not an arbitrary number of days.

If I can leave you with one thing today, it's to never stop learning. Your Google Ads campaigns never stop learning. I have been doing this for a very long time. I'm one of the top experts in this field and I'm still learning new things about Google Ads every week. And so I encourage you to keep your mind open to all that additional data coming your way so that you can continue to learn and grow as well.

And of course, if you're feeling stuck or want a little more one-on-one attention for that learning and growing, you can book a Google Ads coaching call with me via my website, jyll.ca. That's J-Y-L-L dot C-A. 

I'm Jyll Saskin Gales, and I'll see you next time Inside Google Ads.