Intro to Machine Learning (Google Ads)

A2O Digital touches on the world of AI and how it relates to Google Ads and machine learning.  We cover the fundamentals of  bidding, and discuss the three types of bidding strategies.  Tune in to this episode to gain insights into automated bidding and determine if it aligns with your needs. 

Transcript

Copy Transcript Copied

Tim  0:19  
My name is Tim Coleman and I am your host. I am also the managing partner of A2O digital a full service digital marketing agency working exclusively with home service businesses. In today's podcast, we will discuss how our agency helped the home service business triple their lead volume while cutting their cost per lead using Google ads, automated bidding strategies, or machine learning. Our guests today are Ashley Bencsko, and Elainey Yerger. Both members of the A2O digital team, Ashley has been working in the digital marketing space since 2008 and Elainey since 2018. Both are spent a significant portion of their time on paid search, and more specifically, Google Ads. Ashley and Elainey, welcome. Thanks. So let me just get started by by just kind of level setting us all on on Google ads, and PPC, we talk about PPC, right, right, the ads at the top of a Google search results page. And as a, as a home service business, probably at this point, most of advertised on Google and seeing these types of ads. And, and we're used to kind of understanding, it's an auction. And we can bid on a keyword, we can if we're a plumber, we might want to show up when somebody types in plumbers near me. And maybe we're willing to pay $50 for a click when somebody you know, types that in and sees our ad, you know, that's what we meet. That's what we would refer to as manual bidding, right? And this is traditionally how Google has worked. And we got used to it that way. And right, we're going to bid on keywords. And you know, this is kind of great. This gives us an idea where we're coming from this can refer to again, as manual bidding. We know today, we want to talk a lot more about Google's machine learning strategy, their AI. Before we do that, though, so we need to define what we're really trying to get from Google. Right, what we're going to, so let's start kind of with that. And that's conversions. We want folks to take some action, which is why do we want to show our ads when these uncertain keywords right, so let's kind of start there, what is a conversion? What are some of the kinds of conversions that a business owner in the home service industry, like, wants to get from Google.

Elainey  2:56  
I think the most common one is definitely a phone call. Most people in the home service industry, you want to call you want to talk to somebody kind of want to explain your problem that's going on with your home and you know, schedule that appointment. But it also can be any other really type of lead. So if you have live chat on your website, or contact forms, or if you have an online scheduler, that can be a conversion kind of lead and conversion are kind of synonyms. So if you are trying to get in contact with your ideal customer, and you want to get them on the schedule, that would be considered a conversion. There are other ways that you can set up conversions. And you can tell Google that you and there are more top of funnel. So there are people that are earlier in the buying cycle. So this may be something like if you have a download on your website for a buying guide. Those are other kinds of conversions you can have on your site or like an email signup, if you have that. Once you kind of have all of these conversion metrics, and then we can feed them back into the ad account, that's when it opens the doors for these automated bidding strategies. So you, you really want to be mindful of like what conversions are important for your business, and make sure that those conversions or leads are fed back into that account.

Tim  4:23  
Right. So most typically, again, it's gonna be a phone call and that type of thing. A contact form a live chat, like you said, but there are other things you certainly want to be mindful of what you're trying to teach Google to do. So Google's not new to the machine learning space, right? It might be new to you. It's a little bit we've been using it for a couple of years now. But there are automated bidding strategies have actually existed for a long time. But historically, there were limitations to who could use them how they could be used, and how successful they were. Can you guys just sort of explain a little bit about Google's automated bidding journey?

Ashley  5:06  
Yeah, so I think we're talking about it today's makes it feel like maybe this is a brand new feature, when in reality, like you said, Tim, it's existed for a long time, it's been, I think over a decade, they've technically had some of the functionality available. And the journey kind of goes from manual bidding, which we described earlier, where you're truly setting a bid per keyword at auction. There's enhanced CPC, which was kind of a bridge, I viewed as a bridge between manual bidding and the machine learning strategies or automated bidding strategies, because you were manually setting bids, but they were applying themselves a little bit of like changing of bids on the Google side of things. So it was kind of bridging the gap. And there's full blown automated bidding strategies where you are not in the traditional sense. As the advertiser setting any bids for any keywords at any auction, you're giving Google essentially instructions about what you ultimately want them to accomplish. And then they're deciding at the time of auction, what they want to do, or what what they want your bid to be, essentially how competitive they want you to be. So that's kind of how it's evolved, like moving from manual to machine. The information that's available to a machine learning or automated bidding strategy advertiser is largely placed behind the scenes and not visible to a manual editor, right. So if you're a manual bidder, you can see some demographic data about the advertiser or about the searcher, you can see things like maybe what age bracket they fall into, you might be able to see some high level data about income, you may be able to see, male or female. There's some high level stuff you can see. But the machine behind the scenes knows a lot about everyone that's on Google more than we really want Google to know about us. But the reality is, they know everything about us, they know if we've recently been married, if we moved, if we like sports, if we're into cars, or like home and gardening, like do we frequently purchase certain things have we been searching for certain things recently, like they know just everything about us. So when you use the machine learning strategies, you're able to leverage both what you can see manually, but all the stuff that the machine knows about you behind the scenes. So you're truly at a disadvantage as a manual bidder in some instances compared to a like, automated bidder, because at in real time, at time of auction, Google can see all this information about you. Now, the machines know a lot about us for a long time. But I think what's changed over time is how well Google's using or leveraging that data as the machine and then what type of results they're pumping out. So if you're someone that tested some of these strategies, these automated strategies 5, 6, 7 years ago, and you were disappointed by the results, like we saw the same thing. The results it was producing several years ago, were honestly from our experience kind of garbage, it was like, This isn't good. In fact, we're getting worse results than we can get manually, manually, we have control over what's going on. Like why would we sacrifice this control, give it to the machine for it to spit out results that aren't as good. But what we've seen over the last two or three years is that all this information that Google's had for a long time, is being leveraged really well and producing results in many cases that are better then what we can produce as a manual bidder or someone using a manual strategy. And so I think that's where a lot of the evolution is coming, is how well Google's able to use the information that they've had for a really long time in order to produce results. So that's a little bit about kind of like the evolution. The other thing is, is that it didn't used to be available to smaller advertisers. So we work traditionally with medium small size home service businesses. And I remember reading about these automated strategies about 10 years ago and saying, Oh, these sound very cool, I want to test these you know what's gonna happen with these. But then reading at the end of that description, you need to have a minimum of 1000 conversions a month in order to test a strategy. And frame of reference, most small businesses, maybe have a couple 100 conversions a month in paid search, and that's a lot for them, right. So to get to 1000 a month wasn't possible for most people. And Google really needed at that time, more data to even produce results that were remotely positive with the machine. What we've seen there as far as evolution is, they're now recommending that you have a minimum of 15 conversions a month in order to access these machine learning strategies and to leverage them to the best of luck or for them to leverage it to the best of their ability. So the volume of data or the size of business that you need to have in order to use these machine learning strategies have it's really basically available to almost everybody, right? Almost everybody can choose to leverage these strategies which is is really exciting because a large, large volume of businesses out there in the world are of these smaller sizes. So it's really just making it accessible to almost everybody.

Tim  10:09  
So if you're a local business, you jumped into the water seven or eight years ago, the water was probably pretty cold because you're just just weren't given Google enough data. The waters warmed up a little bit. So come on in and test it now. Right, so let's, let's talk about a little bit so we define manual bidding is kind of, you know, you wanted to show for plumbers near me and you told Google you're willing to pay $50 for that, click. Now, we're kinda say in Google, you decide how much to pay, I want conversions. I want folks to call me, you tell me how much I have to pay to get somebody to call me? That sounds like, you know, given the key to the castle to, you know, the wrong person, right? Like, how, how do we trust Google? That seems like, like these automated strategies require some confidence we have in Google to help us achieve our goals. How do we find that trust?

Elainey  11:12  
So we're probably some of the biggest critics of Google, we also question if we should trust them, which is why it's super important to always audit your results, always make sure that you're mindful of like what the quality is coming out of your account. So if you have this conversion tracking setup, and you're able to see your phone calls, and listen to those calls, audit your contact forms, see what the quality is that's coming through these leads, and make sure that they're driving the good results. Because the other piece of this is, if you Google does best when the data is good. So whatever data that you're feeding into Google, it's going to produce out results based on what you give it, it's on the other side of this. So if you give them bad data, they're gonna give you bad results. So you want to make sure that you're always staying on top of things. And you're really like auditing Google, you know, I think you trust them to do it. But don't trust them to set it and forget it and walk away from it, you need to have somebody in there watching it and managing it.

Ashley  12:20  
Yeah, you have to remember with the machine like it really is a machine, it doesn't have feelings, and it doesn't. So like one of the strategies for automated bidding strategies. And we'll talk about more about the different options and in a future podcast, but one of it is basically you tell Google get as many conversions as you can with my budget. So Google looks at that and says, Okay, I'm gonna get as many conversions as I can with this budget, that doesn't necessarily mean that they are thinking about the quality of those conversions. So they may get you 100 conversions used to have 50. Great, they've done their job. They said, We got as many conversions as we could with that budget, we got more than we were getting before. Like, the machine says, like, check, we did our job, right? Where we don't necessarily trust Google, or where I think no one really should is what Elainey said. Are those 100 conversions of quality? Did they result in revenue? Did I you know, were they calling for services that I provide, or that I'm attempting to target, that's where I think it can get a little blurry and you have to not trust the machine. So you do have to hand over some amount of trust, you have to turn it on, you have to start testing, you have to start evaluating, but that evalutating piece is the most important thing I think is just because it says it went from 50 to 100, doesn't mean it went from 50 conversions to 100 conversions of the kind that I want are the kind that produce revenue for my business. That's where you want to make sure that you get involved. And remember that the machine doesn't have feelings. So it's not looking at those and saying, Oh, I feel really good about these and really bad about those. It says you told me to get more, I got more. And so I think that's where you really have to kind of get involved with the machine.

Tim  13:59  
And nothing's changed there. We've always had to audit and assess results at periodic levels. There's there's no difference here, we have to audit and assess to make sure it does help to have some results from a manual campaign to compare obviously, but ordeauditing and assessing is the way you're gonna have confidence, I think at any, any lead source. So can you give us a taste of what results we've kind of seen using an automated bidding strategy?

Elainey  14:31  
Um, so in one of the early accounts that we really tested an automated bidding strategy and it's a large metro area, extremely competitive, which makes it really expensive. Cost per clicks are really high. It was really hard to kind of play the game with manual bidding. And on top of that, they were limited by budget, which kind of makes sense because it's a really huge Metro area, it's really easy to in situations like that for spend to get out of control. Over the course of a few months, we rolled out conversion tracking. So we were able to get that lead information that's been flowing out of Google, we've written now flowing that back into Google, so we can tell exactly which keywords are driving conversions. And then we turned on an automated bidding strategy. Within the first few months, we saw conversions triple, and it reduced cost per conversions by 60%. So that means that for the same amount of money that we were spending, we're getting three times as many conversions three times as many leads, but they're costing us a lot less. So that's the exciting results is when you know, the goal is to get as many, many leads as possible. And we were able to, you know, deliver on that.

Ashley  15:57  
I think the fun thing to highlight there too, is it wasn't like least tripled from one to three, right, which is tripling, but not as exciting, like least triple from like 100 to 300, right. So like, approximately, that's give or take, I'm like cost cut from like 300 to 100. So like real numbers, not like even though you know, percentages can be deceiving, but you're really seeing like huge increases in the volume of leads, and really significant decreases in that cost per lead. So both are a huge one.

Tim  16:25  
When you guys look at our customer base, now, what percentage of the lead, what percentage of our customers rather, are using approximately a machine learning strategy or an automated bidding strategy versus manual.

Ashley  16:38  
We're approaching like 75% are using some some degree of machine learning, even if they're not implementing it into every single campaign for a variety of reasons. We also tend to treat like branded terms versus more generic terms like plumber versus, you know, ABC plumbing differently. And so but to some degree, about 75% of the businesses are using some version of a automated bidding or machine learning strategy and their Google Ads account. And I think one of the biggest hurdles to getting there, too, was that trust component that you you mentioned earlier, Tim, you know, handing over that control, giving it to the machine just feels uncomfortable. They've seen costs for their own businesses rise over time in advertising for Google ads. And so it's like, should I hand over? Should I hand over this control? And really, it's it's been results that have helped make the conversations easier. So, you know, if you're not doing it, I think we would advise at least talk about it with your marketing representative at least have a conversation? Is it a fit for your business? If it is why there's a variety of strategies you can try, which as you said, we'll talk more about in another podcast, but like, if, if it is a fit, which strategy is the best fit? How should we be managing our budgets? What conversions should we be tracking Elainey mentioned earlier, like top of funnel versus bottom of funnel, the machine tends to do better when you tell it to go after conversions that are bottom of funnel, so really close to the actual purchase. Because you're essentially allowing Google to hone in on your ideal customer, right. So if you hone in on those conversions that are close to the bottom of the funnel, either at purchase or right before purchase. That's what your target customer looks like. That's the kind of person you want more of. And so that's kind of the conversion you want to define for Google, and then give them the instructions of getting more of these, right. So but we would definitely recommend, if you haven't had the conversation, at least start with a conversation. Sometimes it takes some time to get acclimated to the idea of giving this control over to the machine to understanding how this process works, it's a little more complicated, then I pick a keyword, I pick a number for my bid. That's what Google has, that's what they can use great the machine, it's a little more like, fuzzy as to what's happening, but start the conversation and definitely encourage testing. And also, I would say, don't be discouraged if your initial round of testing doesn't give you the types of results that you're hoping for. Sometimes it takes a lot of adjusting to a variety of factors within that strategy to get your optimal results. And so just stick with it for a little while before you abandon it. There is nothing worse than have when we work with a customer and they turn on one of these strategies. And we're very excited because we think it can produce great results. And then three days in, they get nervous and turn it off. The machine hasn't even had time to learn what's you know what's happening and what it should be producing. So you also want to think with this type of approach that I have to allow some time at least a couple of weeks, maybe a couple of months before I say this is working or this is not working.

Tim  19:49  
Yeah, okay, that makes sense. And it doesn't have to be an either or thing right. You can test it in some campaigns and not others kind of dip your toe in the water, right?

Elainey  20:01  
Absolutely. And we kind of encourage that. Because if it's an all or nothing and knock on wood, if something went bad, then like, let's not have it impact everything. Let's have it silence to whatever we're testing on.

Tim  20:14  
Yeah. And I think I think that has that has helped us. The fact that it doesn't have to be an all or nothing strategy. And we can kind of walk into some automated strategies rather than like, you know, diving in the deep end.

Ashley  20:30  
Yeah, I think one more thing to highlight, which is just helpful to know, as a person advertising with Google. Google has not announced anywhere that they're going to take away like manual bidding strategies. Definitely never, I've never seen that publish or anything like that. But what I do know is, every time our agency works with Google, have conversation with Google, their focus is migrating people over to these automated bidding strategies. And again, like, they've never said that they're taking away manual bidding. But I think what I always question to my brain is the writing on the is the writing on the wall for manual bidding at some point in time being pulled from like, the list of strategies that you can utilize. And maybe it will happen, maybe it won't, but you want to at least have an idea of how these strategies work so that if you get to the point where you can't use manual bidding, you kind of have an idea of what's available, you know, how you would structure it for your business? What would you know what maybe legwork do you need to do to get these conversions trackable, just tracking phone calls, just because you you know, say I want to track phone calls doesn't work, there's there's a process for making sure that you can track your phone calls if you're not doing that already. And and counted as a conversion. And same thing with if you're using live chat or contact forms, like there's a process to put that in place. So you want to at least start kind of thinking in that direction. I think as a business, you know, savvy, I'm looking forward at least a little bit. Even if your best results still come from the manual bidding strategies, you really want to understand what else is out there. Because I would personally not be shocked if at some point in time in the future, they said, you know, we're sunsetting manual bidding on this date. Again, not saying that it's happening, not saying that it's been published, but we just we It feels like the writing's on the wall to a certain degree.

Tim  22:13  
Yeah, for sure, so Elainey, Ashley, thank you that was insightful. A lot of information about automating bidding strategies with Google and as a business owner starting to think about that. Thank you for joining us today. I appreciate you guys spending some time to listen to this podcast. Ashley and Elainey, alluded to different types of strategies in automated bidding. We're going to take a deeper dive into that and in our next podcast, and I will continue with water based analogies, I guess as well, since I've used so many here today. And, we'll hopefully see you there. Thanks and talk to you soon.