Inside Google Ads podcast: Episode 84 - The Algorithm
What do Google Ads practitioners mean when we say “the algorithm”?
You've heard it before…
The algorithm needs time to learn.
The algorithm needs sufficient data.
The algorithm doesn’t like bid limits.
The algorithm, the algorithm.
How does the Google Ads algorithm actually work?
An algorithm is just a process to follow, a set of instructions or rules. In Google Ads, there are many different algorithms, most of which are now powered by AI, and they determine all aspects of your campaign performance given your inputs.
Today, I'm answering your burning questions about the bidding algorithm, the keyword algorithm, and the Performance Max signals algorithm. Are you ready?
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 84, The Algorithm.
Our first question comes from LearnEnglishWithMe852 on YouTube, and they say, how do I get my bidding strategy out of learning, fast, in my Google Search Ads?
Whew, that's a tall order.
You may have noticed when you start a new campaign, when you change your bid strategy, or when you make another big change to your campaign, like changing all your keywords or completely redoing your ads, that suddenly your bid strategy is learning.
This status actually means a lot less than you think. That “bid strategy learning” status will be there for five days. Whether you spend $5 a day, $5,000 a day, or $5 million a day, it will still be there for five days.
What actually matters for your bid strategy to learn is that it has enough inputs to determine what helps you achieve your goal and what doesn't help you achieve your goal.
For my answer, I'm going to focus on Maximize Conversions bidding, but the same principles apply to whatever it is your bid strategy is trying to achieve, whether that's conversion value or clicks or CPA, et cetera. Though of course this would not apply to a manual strategy.
So with Maximize Conversions and Target CPA, you've probably heard people, including me, say you need 30 in 30, 30 conversions in 30 days, in order for it to work.
Why is that? Well, when you're getting about a conversion a day on average, that's typically the bare minimum amount of data needed for Google Ads to determine who is the right kind of user, the right kind of query, the right time of day, the right day of week, the right location, the right audience, etc., vs. which of those aspects do not work well for your campaign. And that rule was devised expecting you to spend $100 a day on Google Ads or less.
Now, if you spend a lot more than that - for example, I just audited an account where they spend millions of dollars a month on Google Ads, so for them, their campaigns only need a day or two to learn. And then by then, they have hundreds of conversions and they're ready to go.
Conversely, I've audited an account recently that has like five or six conversions a month in their Search campaign, so it's going to take three to four months for Google to figure out what works well and what doesn't.
So how do you get your bid strategy out of learning fast? Give it a ton of conversion data. Usually that means increasing your budget, but using something like micro-conversions can help you get out of that learning period faster if your conversion volume is low.
For example, if you're an ecommerce store and you're looking to drive purchases, but because of your low budget, you only get a purchase every couple of days, maybe you want to start optimizing for “Begin Checkout.” It's possible that for every three “Begin Checkouts,” you get one purchase. So optimizing for “Begin Checkouts” gives you a ton more data for your bidding strategy to learn.
Episode 39 of this podcast was all about micro-conversions if you want to learn more about that.
Finally, I'll say that getting out of learning fast shouldn't be your objective, I don't think. It's not like when things are in learning, they're not working well, and then when they're out of learning, they're suddenly working well. Your algorithm is always learning. So if market demand changes, when you make changes to your campaign, if something happens on your website, all of that is going to be taken into account as your bidding algorithm sees different dynamics happening.
Like humans, your algorithm is always taking in new information, learning from that, and adapting. Well, at least that's how I try to be as a human. I hope you do, too.
If you're still finding this stuff kind of confusing, you might enjoy my course Inside Google Ads. I have more than a 100 in-platform tutorials, plus you get to ask me your burning questions every month at an exclusive one-hour call just for course members. You can learn more at learn.jyll.ca, that's J-Y-L-L dot C-A, or follow the link in the episode description.
Our next question comes from Evan Weber on LinkedIn. This is when I shared on LinkedIn something that I talked about on this podcast recently, that when you have Broad Match keywords, I would look for your Search Terms Report to have about 80% relevant search terms, which suggests Broad Match is doing its job. And Evan said, 80% relevant. You're kidding, right? One irrelevant click per week is too much. And then Evan and I went back and forth again in mostly polite disagreement. Thank you, Evan. And then he said, the goal is to have a 100% clean Search Terms Report. I hope you would agree.
Evan, I respectfully disagree, and here I'm going to expand upon why.
Broad Match keywords are designed to find as much relevant traffic for your business as possible. Now, of course, on day one, they don't know what's relevant and what's not. They need your conversion data to figure out what are the right kinds of queries for your business and what are the wrong kinds of queries for your business. Even when they figure that out, the whole point of Broad Match is it's constantly going to be testing new things, looking for novel queries, looking for long tail queries, just to test and see if we might find some good performance there; perhaps because other signals are in place, like the right audience or the right location, where typically you found good performance before.
So because of that, when you're using Broad Match, no, a 100% clean Search Terms Report is not the goal. In fact, that suggests that you've probably just added so many negative keywords to your Broad Match campaign that it's not even Broad Match anymore.
And I get it, there are a lot of times where we want or need to have control. In that case, you're going to want Exact Match keywords because they will give you that much tighter control. And when you're using Exact Match, then yeah, I am looking for a 100% clean Search Terms Report.
But with Broad Match, you probably can't even see half of your search terms anyway because of the hidden “other search terms.” And for those you can see, I believe it's unrealistic to expect every single thing to be perfect because that would then mean that your Broad Match algorithm is not testing, is not bringing incremental traffic, is not doing what it's designed to do.
That would be like going to cut an apple and instead of reaching for a knife to do that, reaching for a lightsaber. Okay that could work, but it's not the best tool for this job, even though it can be a great tool in other circumstances.
Now, once your Broad Match algorithm has had more time to learn, it has a lot of conversion data, it's run a lot of experiments on its own and figured out the kind of stuff that works, the kind of stuff that doesn't work, then yes, you will probably see much more relevant search terms than you did on day one. And this is actually a mistake I'm seeing a lot of people make right now with AI Max. They'll turn on AI Max and after two days say, “Oh my gosh, it brought in the most irrelevant search terms. That was garbage” and turned it off. The AI Max algorithm hasn't had time to learn yet! These things need time to learn, to collect data, to test it out, to see what works and what doesn't so they can then optimize for you.
That's why for people who are brand new to Google Ads, I say unless you're prepared to spend about $1,000 a month on ads for at least a few months and potentially see no return on that investment, you shouldn't invest in Google Ads.
Similarly, if you're not prepared to test Broad Match or AI Max or Demand Gen or PMax for at least two to three months and potentially see no return on that investment, then you just shouldn't test them in the first place. Stick with your tried and true strategies.
You can't have your cake and eat it too. Either you want to grow and expand the business, which means you have to allow for some budget that's not going to end up with great results, or you want to stick with the tried and true strategy that works for you, at the risk of not growing your business or not growing your client's business.
There's no right or wrong way, but that keyword algorithm is going to end up making some poor choices along with some good choices until it has sufficient time and data to learn on your business.
If you're not sure whether testing one of these new features is the right fit for you, or if it's time to give up on an experiment you're already running, you should book a call with me. These are exactly the kinds of topics I tackle with my Google Ads coaching clients, whether you're a small business owner who's trying to DIY your ads yourself or an in-house marketing team that's just looking for an outside perspective. You can book a call with me by visiting my website at jyll.ca. That's J-Y-L-L dot C-A or follow the link in the episode description.
Our final question today is about signals and Performance Max. And I actually got two questions from you recently about this.
First from Joran Lamisse on LinkedIn, they say, PMax is keywordless, but what about Search themes? Isn't it a hint that Google uses for Search targeting?
And then K.Nachrashid on YouTube says, what about the asset group signals in a Performance Max campaign? Why don't they work for targeting the type of audience you want to target?
Repeat after me: a signal is a signal. It is not true targeting.
A signal is a signal. It is not true targeting!
Audience signals and Search themes are some of the best branding that the Google Ads product team has ever done, in my opinion.
Remember what I shared in the first question around your bidding algorithm? In PMax, you only have Smart Bidding. You only have conversion-based bidding, and so that means your Performance Max campaign will be singularly focused on achieving your conversion goals, whether that's getting as many conversions as possible, as much conversion value as possible, hitting your desired target cost per conversion, or your desired target return on ad spend.
Your PMax targeting algorithm, which determines the right channels for your campaign and the right audience and content targeting for your campaign, does not care what you think it should do. It just wants to achieve its goal. So feel free to add audience signals. Sure, it'll test those out, but it's only going to truly target those if they drive good results. Similarly, feel free to give it Search themes. Sure, it'll test those out, but ultimately, if it places your ads in Search or Shopping at all, it's going to do it for the queries that it learns are most likely to achieve your bid strategy objective.
Your bidding algorithm is what guides your targeting algorithm. That's true for PMax. It's true for AI Max. And I'd argue that's also true for Shopping. Shopping has always been keywordless. It uses your product feed as its guide, but ultimately, two Shopping campaigns with the exact same targeting - but with one on Maximize Clicks and one on Target ROAS - they're going to do completely different things because they have completely different bidding objectives.
Does this mean that you shouldn't give your Performance Max any signals? No, it can't hurt. But once you set up your Performance Max campaign, you give it your audience signals, you give it your Search themes, I would not recommend that you bother trying to fuss around with those later. Once your targeting algorithm has learned from your bidding algorithm the kinds of users that are converting in PMax and the kind that aren't, it's going to stick with what it has learned.
So the reason to have multiple asset groups in a Performance Max campaign is not because you want to have the same creative with different signals. It's because you want to test out different assets, different creative, which may potentially resonate with different audiences. That's why they're called asset groups and not audience groups.
The key thing I want you to take away is that you do not need to be intimidated by the algorithm. Just like you need time and data and practice to learn how to use Google Ads effectively, Google Ads needs time and data and budget to learn how to drive the best results for your business. If you can't be patient and give it time to learn, then it can't give you the results that you need.
Today's Insider Challenge is this. An algorithm that we didn't talk about today is the ad creative algorithm. Let's say you have a Search campaign on Maximize Conversions and you want to optimize your headlines and descriptions. How would you go about doing that? How would you know if the algorithm has had sufficient time and data to learn?
The beauty of the Insider Challenge is there's no right or wrong answer, just an opportunity to stretch your brain on real life Google Ads problem solving.
Last Episode's Challenge, Episode 83, was this. How would you go about finding the right YouTube placements for a video campaign?
There are many, many ways to do this. Here are three of my suggestions:
You can check your own YouTube analytics to find referring videos and channels. So you can see which videos or which channels are sending people to your videos. It’s a great place to start.
I recently did a sponsored video with Similarweb, full disclosure, so you could use Similarweb to find other websites and brands that your audience is interested in and then look up their YouTube channels. I show you how to do exactly this in the Advanced Audience Targeting with Similarweb video I created. I'll include the link to that in the episode description.
If you have existing PMax or Demand Gen campaigns, check what placements you're already on for both placement ideas and placement exclusion ideas.
And then bonus tip, I still prefer audience targeting. I care more about the individual than the channel. I find that placements aren't very scalable and they can get really expensive.
Also note that you need to choose at least 10 placements for your Video campaign to be eligible to run. But hey, if you want to reach Google Ads practitioners, go ahead and place your ads on my YouTube channel.
I'm Jyll Saskin Gales and I'll see you next time Inside Google Ads.