Unknown Speaker 0:10
Our next speaker is one none you’ve probably ever heard of before, we thought we’d give him a chance because he didn’t have much experience to see how this goes. I’m just kidding. Welcome, Fred. So Fred has almost as much experience as I do with this, like, I’m now on 23 years. What do you have to Fred? Ah,
Unknown Speaker 0:29
depending how you counted back to 1998. When I first started using, go to, yes, so
Unknown Speaker 0:34
we’re at both at 23 years. So we’re showing the gray hairs. Well, actually, Fred was you must dye your hair.
Unknown Speaker 0:44
I’m getting their bread.
Unknown Speaker 0:46
Yeah, yeah, I
Unknown Speaker 0:46
understand.
Unknown Speaker 0:48
I’m just letting it rot. So anyway, I’m gonna stop talking. Now. Everyone, this is Fred. If you don’t know, Fred, you should. Fred’s the CEO of optimizer worked at Google for a while previously, starting a paid search because he was a photographer back in Belgium. I’m remembering my history. Alright, so I will let you take it away, Fred. Let’s wow was with some fun stuff.
Unknown Speaker 1:11
Thanks for admitting. So 25 minutes to talk about smart bidding. That’s like a tall order. There’s so much we could talk about. So I’m not going to spend too much time on myself. But really, this whole session is about making the most of automation, when most people actually think that you put smart bidding up, and you’re sort of done at that point, right. So my background, 10 years at Google helped build AdWords helped acquire merchant, which is now Google Analytics. So a lot of great stuff shared by Brook, actually a lot of stuff, I’m going to refer back to as well on that to save some time here. But really, when we talk about smart bidding, and when we talk about Google ads, I think that you can think about it like a child like a person right? Back in the day, when we were working on the ad system, when it was still called AdWords, it was really like a child, and you needed to do every single thing to make sure that that child would still be alive The next day, right, you had to feed it, you have to help it poop, you had to help it sleep. Any of you who have children, you know, nothing is automatic with a child. And so everything you do is manual. But now, a bunch of years down the road, Google, or AdWords has grown up into what’s called Google ads. And it takes care of all of these details. And if you think about smart bidding, it’s kind of the same, right? You had to take care of keywords, bids everything manually. Now, those details are handled, but what you can still do, hopefully, is influence that child, Google ads to do the right thing for you. And dad, by the way, is not automatic, it doesn’t automatically do the right thing, you have to help guide it to do the right thing. by teaching it. All of this is based on machine learning. And as the name machine learning says, learning requires a teacher and we can still be the teacher to that system. So that’s what I want to share with you today. Now, as far as smart bidding, right, I know a lot of the experts tend to be a little bit hesitant to adopt smart bidding in many cases. But I just wanted to put out a number there that more than 70% of Google ads, advertisers are currently using some form of automated bidding. So in many cases, that would be smart bidding from Google. And this site has all the way back to 2019. So it’s probably a higher number at this point. Now, I also wanted to share some numbers that we pulled from the optimizer ecosystem. And so we looked at a sampling of a few of our customers. And we looked at what was the conversion rate between those using automated bidding, and manual bidding. So I’ll break it down into three groups here. So on the higher end of spend, right, and how you run it, it’s all relative, right? But kind of looking at buckets of spending 15k a month and up, there is actually a very decent lift in smart bidding conversion rate performance. And now we could go and argue like, should we look at conversion rates? Or should we look at profitability, and we’ll get into that in a bit, right, but just we need to look at one metric here. So let’s say it’s conversion rate, while conversion rate gets better, with smart bidding. And I think that’s probably because machine learning loves Big Data loves lots of data. And big accounts tend to have lots of data for the machine to find little signals in and actually help you get more from from those accounts. Now, on the flip side, you got small accounts, small accounts, and these are the ones spending under $1,000 a month, they can be challenging for agencies, because there’s often not really a ton of data data for you to make good decisions. So if you’re doing manual bidding, and you’re trying to do geo bid adjustments, or device bid adjustments, or audience or whatever bid adjustments you’re trying to do, even straight up conversion rate across all the clicks you get, well, if you are a law firm and you spend $1,000 a month like you’re getting what 10 clicks, you’re not going to make really good decisions on 10 clicks. That’s not a representative sample, right? So again, in that case, machine learning does for a reason I can’t necessarily explain but it does do Better than if you were trying to manually manage those accounts. And then interestingly, in the middle of that data set, where you have kind of the average spend accounts, there’s really not that much benefit that you see from using smart bidding.
Unknown Speaker 5:15
Also interesting to see is that these accounts with manual bidding tend to already have the better conversion rates. And that’s probably because there’s that right balance have enough data, the account is big enough, the agency, the account managers, probably spending a decent amount of time putting in place some decent strategies. So by layering on top automated bidding, you get a little bit of benefit here and there, but it’s not as clear cut as in the other two ends of that spectrum. Now, if you’re going to do smart bidding, and many of you probably already are, but if you’re not, or if you’re seeing some issues with it, I want to point out some prerequisites. So first of all, and Brooke covered this really well. But it’s about selecting an attribution model. Right. So the first point and the key point is do not use last click attribution. And you’ve heard me say this before, if you’ve attended one of my sessions before, but the point is simply this, if you think about conversions in a fairly simplistic way, kind of the old way, but you think of it as a funnel, while you basically build awareness at the top, and then eventually you get people to know what it is they want to buy. And you, you sell it to them through a very specific keyword. Last click attribution gives all debate all the value from that conversion having happened to the last event, and it basically ignores everything happening at the top. Now, us humans were smart enough to recognize those keywords that are important to our business and that operate at the top of the funnel. The machine is dumb, right? We call it smart bidding, but it’s actually really dumb, it just looks at the data that you give it. So if you give it data that says, we get no conversions directly coming from upper funnel activity, it will decide to bid down for all of that. And it basically kills your funnel, the funnel gets narrower and narrower. And eventually you kill off your traffic. Right. So that’s why you can do last click attribution if you’re going to be using smart bidding. Now, what do you do instead? Well, the two most straightforward options would be to use the position based or a time decay model. As we see those on the right, the time decay model is the closest one that you currently have to a last click attribution. So that’s a very easy and nice one to switch into, because you’re not going to see that much change, but you’re still going to assign some weight to the early activity. If you have enough data, then obviously use data driven, because that’s the fancy machine learning one. Now the second prerequisite, again, Brooke kind of touched on this, but it’s about setting goals and be reasonable. Right? So we’ve all heard a story of the agency that got a new client, and the client said, our target row as is 500%. And so for six months, that agency is operating at a 500% margin and kind of looking at the numbers and saying, well, we could have driven more profit for you. But hey, declines at 500%. While then they go to the client, and the client says, Well, the only reason we said it needed to be 500% is because the last agency did 400%. So we figured let’s raise the bar, right? But that’s stupid. Because if you look at this chart here, if you have a high target row, as you’re basically killing your volume, you’re focused on margin on profitability, but that’s the percent right. Right, I might make 500% on my one click that I bought, because I was so limiting in how many clicks I could back. On the other end of the spectrum, if you set a really low target role as well, yes, you’re going to maximize revenues. But at some point that cuts into profits. And this chart explains that right? So if you have a low target role as you’re going to have not that much profit, because you don’t have very many sales. If you have a high target row as you’re also killing your profit. So there is that point on the curve, where it starts to dip, right the the apex of your curve, basically, that’s the right point of target row as that maximizes your profit. Now the problem is, that’s often not an obvious point to find, and you have to experiment together. And then thinking back about bid management, right? Well, we all were setting manual bids, how did we go about that, while we would often set a bid, we would go back and look at the account the next day, and we would see what the performance was. And we’d make a course correction we’d slightly adjusted bid. It’s really no different. Oftentimes, when it comes to smart bidding, you have a target. But the target has to adjust based on what you see based on the profitability based on the business goal you’re trying to achieve. So that’s the second prerequisite is have a sensible target and also have a unique target. Right too many people say, Well, I
Unknown Speaker 9:31
want like super high rollers, but I also want to maximize my profit. And then they got three other goals and it’s like all convoluted and that just doesn’t work in PPC, right? You have to be specific about your goal. Now, once you think about working with smart bidding, right, let’s actually go back to the days of my daughter at least with a chocolate all over her face. Right? Why? Why were we doing all of these things. So in AdWords in the old day, you would pick your keywords, you set your bids, you do your bidding, adjustments, you’d set up each single ad. And you were managing all of this for the purpose of getting more leads and sales and signups. And I don’t know if any of you watch the show Lucifer. But he asked the question is, what do you really desire? Right? And that’s a good question. Because Do you really want leads and sales? Like I run optimize, right? I mean, I love leads. But do I really care about leads? No, I want people to pay for the software. I want subscribers, right? So once you start asking that real question, what is it you’re really trying to achieve? While I’m paying clients, I want profits, I want revenue. These are the business goals. And somewhere along the way, we tend to lose that connection. Because we were so focused on the left side, right managing the detail. Now in the virtual world of smart bidding, we can flip that upside down. And we can actually say, Well, the thing we should optimize, the thing that I should work on, is figuring out how to better identify the value to my business and communicate that back into this machine learning system. And then that machine learning system will handle the detail. And I can still put some controls on it, I can still monitor it, I can still put up alerts. But fundamentally, I’m shifting how I think about doing Google ads, it’s no longer about me really caring about the keyword. And we all know about the match type changes that happened about two weeks ago, right? And those bits us off because we lose control. But at the end of the day, if we’re using smart bidding, and we’re letting Google figure out the right keywords, well, they’ll find the right price for those things as well. And if they don’t find the right price, it’s often because we haven’t done a good enough job of communicating what it is that we truly care about. As far as results go. Okay. And then if you get into smart bidding, the other thing a lot of people don’t realize, and this is up on our blog, but bid adjustments start to behave really strangely. Many of them simply don’t work. But then ones like device type adjustments, they will actually adjust your target CPA, but they don’t do anything for target robots, right. So we have this whole table published that you can take a look at to really understand if I’m going to do bid adjustments on top of smart bidding. For the most part, they don’t work. But if they do work, what is it, I should expect them to do? So take a look at that. Now, there’s really three levels of sophistication. And there’s more than that, but we’ll start with three levels here of smart bidding control, using conversion tracking. So the first level is really saying, Okay, I’m gonna do a good job of tracking my conversions. And that, sadly, is not a given. I mean, I still run into advertisers every week, who say, Oh, well, but conversion tracking was broken, or we know we set it up, but we don’t really know that it works. And so it’s like, Okay, well, then how do you manage your campaign? Well, it’s fine if you do it manually, because in the back of your head, you’re kind of looking at the bottom line, and you probably understand what is coming in conversions wise. But that’s tricky for for a smart bidding automated system, it does not have that insight into the business unless you give it that insight. So level one is really fixed up your conversion tracking, and put into things that you really care about. That’s what you should start at, that’s going to enable you to do target CPA smart bidding. So let me give you an example here of what could happen. So say that you have sales, and we’re going to do a Gao example. So you have sales happening in two different countries. Okay, and all you’re reporting is conversions. And so there’s a default value that goes in with those conversions of one, one per conversion. So for these two sales are these two leads that you get, they both get the same value, so you get value of two, you get two conversions. Okay, so now you put smart bidding into play. And all of a sudden, all your conversions shift to one of those locations, and the other location gets ignored.
Unknown Speaker 13:31
And why is that? Well, that’s because it’s just going after the cheaper conversions. Now, the cheaper ones to achieve, because it’s a lower cost market, are those really deletes that you want it to get? Or are those leads also like not as valuable to you? Well, if you’re not reporting value, and you’re just reporting conversions with no value attached to it, the machine doesn’t know any better, it doesn’t know you have a preference for one location versus another. So just does whatever is most efficient for it. So that brings us into level two of smart bidding optimization. And that’s where we bring value into the mix. Okay, so now we’re really thinking about, well, what is the value of a lead? What is the value of a sale? When we think about a sale? Are we talking about the revenue? Or are we really interested in profit? And that changes the value that we put back into the conversion tracking mechanism? Right, so that’s level two of the optimization. So go back to the GAO example. And now we get to sales happening in different countries. But we’re also reporting that the ones in the UK tend to have a higher value than the ones coming from elsewhere in Europe. Okay, so now, we’ve given a little bit more granular data to the smart bidding system. So what it does now is it says, Well, actually, we’re going to prioritize the sales happening in the higher value country. So the chart on the left here that the globe on the left should have shown the dark in the UK as well, but basically saying everything is now shifting to the UK because that’s where we think we have more desirable conversions. Now there is an element, I’m not really going to go into too much here, it might actually still be that you get your private primarily your conversions from the continent of Europe. And that could be because yes, they are less valuable conversions. But the clicks are also so much cheaper element row spaces, it is actually a better place to advertise, right. But that sort of optimization, the system cannot do it. Unless you give it value data, it knows the cost data, but it doesn’t know the value data until you give it to it. And then it can start to optimize for that. And then the third level of optimization is going beyond the straight up value of what happened in that one moment in time, when a quote unquote conversion happened. And now we’re thinking about lifetime value. Now we’re thinking about omni channel, store visits. You know, there’s even cool ways that you can use this and Brooke was referencing it. But basically, if somebody comes to your website, and they don’t convert, and they don’t do the thing that you really, really want them to do, which is buy something, but they sign up for a newsletter, right? That you that’s still a micro conversion, you could still send in some signal to Google that says we actually, you know, out of all the people who didn’t convert, we like the people who didn’t convert, but sign up for the newsletter better than the ones who didn’t sign up for a newsletter. And that tells smart bidding. Once again, that is like this might be a new customer, this might be something valuable to your business that down the line you’re going to be able to monetize. And so finally more of that traffic. So and Google says, but basically, as all an advertising becomes more automated, we believe that the KPI that you asked the machine learning to go after, and the data that you share with it, these are the optimizations that you need to do, right, so stop thinking about which keyword which bid, but think about how you inform the system to be better at its job, because it’s really good at doing its job. Right, with some monitors in place, but it’s really good at it. But you can only do as good of a job as you tell it to do. By the way, there was this whole thing when I worked at Google, one of the 10 principles of innovation for Google was share all the information with everyone. And this was a human principle. It was about you know, my colleagues, my coworkers, they’re all smart people. If I tell them all the data that I have to make decisions, and I tell them the correct goals of what we’re trying to achieve as a company, well, then I don’t have to worry because they’re smart people, they’ll figure it out. They may not go to the use the exact same methodology to get to the right answer. But we’ll all end up in the same place because we have good information. And think of machine learning as that super smart colleague, who you just need to tell what it is you want it to do so that it can do its job. Now, I told you that smart bidding and bid adjustments don’t work together, right? Well, actually, that was a lie. So if you think about it, what you can do is you can do a rule based logic and say if I get two conversions, and they’re exactly the same, the person bought exactly the same amount of stuff, but they were in different locations. But I know in the back of my head, I prefer a customer from one place versus another. Maybe it’s because I think there’s a different lifetime value. Maybe it’s because I know something about the business that says this is going to be a better customer down the road than this one, even though they bought the exact same thing. Well, I create a rule, and I do a little bit of manipulation. And I say, okay, the value of this one was slightly higher, you tell the machine learning system that it’s effectively like you’ve made a bid adjustment on top of smart bidding, because now it’s going to go look for the thing that was more valuable.
Unknown Speaker 18:22
So you think you may think you lost control. But in fact, oftentimes you haven’t. Now, when you think about how to manipulate values that you put into the system, think about the things that the machine just cannot know. Right? Don’t do it based on conversion rate, don’t do it based on current value, because that’s the stuff that should be in your conversion tracking, the system knows it’s going to work on that. But stuff like Hey, does the system actually think about how to better impacts my business performance? does it know that when I put out a press release, behavior changes, while all of that that’s an unknown, so what I can do is I can manipulate the value or I can manipulate the targets in response to that, right. So I’m kind of still optimizing smart bidding in very unique and maybe sort of backdoor ways, but still getting what I need to do. So let me give you a couple more examples of smart bidding optimization. Right. So there’s this whole notion of offline conversion import. When I thought about that, originally, I mean, it says offline conversions. And I was like, well, that’s for sales teams. That’s for people who have brick and mortar stores like stuff doesn’t really convert in the online world. But actually, that’s not true. Think about it this way. So let’s say that you have two scenarios in the lead gen example. Okay, so the user fills out the Contact Us form on day one. Both of those, you give it a value of one because delete form was filled out over the course of the next couple of days. If you haven’t heard back, you can decrease that conversion value for the person who ended up not being such a good lead. And for the person who does sign up, you increase the value. Right now you can do this in lead gen in this example, but even in e commerce, Do you see the person navigating more pages on your site, looking at more pricing pages, maybe looking at the sizing chart for clothes, if you know those are good signals of a likely intent to buy, in almost real time, you can manipulate the offline conversion import, you can send new data to Google that says, Oh, well, actually, that user is trending towards doing what I want them to do. The Smart bidding system looks at that. And it says, Okay, well, again, everything else being equal, this is the kind of person I prefer. So this is the kind of lead customer that I need to go and prioritize. And so it uses it signals to go and find that. Now, here’s another example blocking competitors, right. So sometimes you work with a sales team, or maybe for some reason, weekend conversions are really not that good. Smart bidding might reduce bids during that time, but that lets your competitors swoop in and take away your upper funnel traffic, when eventually, you know, your sales team would have come back to the office. And they would have closed those deals. Anyway. So what we’ve seen it seen from this from Andrew Luck at savvy revenue, but he basically runs two campaigns. So he has a campaign structure that’s informed by business goals. And he says, during the weekends, I’m going to do manual bidding, so that I still show up, and maybe you have a lower budget. And then during the week day when everything’s good, and it’s all normal, that’s what I’m going to do smart bidding, right? So that’s another optimization that you can do. Now, this was a Brooks example. So she was talking basically about micro conversions. Right. And, and, and sort of similar things of behaviors on the site. She was talking about brain goes into the valuation model, so that you can go to a client and say, Well, look, instead of 300% ROI, alas, we have 318% robots, and you’re influencing she said, the decision maker to give you more budget for next quarter? Well, you can do exactly the same thing. But no, you’re not influencing the decision maker, you’re influencing the bid management system. You’re saying basically, okay, I think a request for a demo is more worthwhile to me than all of these other things. So I give it a bigger value. Now, don’t let a lot of people they get blocked at this stage. They’re like, well, I don’t know what a demo is worth, like is that $10? Is that $20? That’s not the point, I haven’t put values on here. All I’m saying is make them relatively different, right? 123. Okay, three is bigger than one. So that is the thing you prefer, that’s enough for the machine to know that that’s a preference that’s going to influence it to look for the right thing. But be careful if you put in a value of 123. But all your clicks, costs $10, the machine is actually smart enough that it says oh my god, this is horrible return on adspend. Right, I have to spend 100 $1,000 to get one conversion. And Brooke just told me that that was a $3 conversion for $1,000 of ad spend. That’s horrible. And it actually throttles it holds back on traffic because it thinks you’re going to be relatively unhappy, right? So it might have been better to set relative values of 100 200 300,
Unknown Speaker 22:55
still relatively the same. But now we’re in the realm of where it’s more in line with the click cost. So I think I’m Exceeding my time here. I had a bunch more slides. But you know, we’ll do another presentation on this one. I do want to leave you with maybe this last point here. campaign structure, a lot of people say, well, Google tells me that thanks to smart bidding, I need to have a flat campaign structure. I just need one campaign. Basically, Brooke mentioned that you should have multiple campaigns with different target requests based on margins. But here’s the interesting and the most important thing. Smart bidding looks at your conversion as the main driver of how it builds the machine learning model, and then it uses the factors from the query from the auction. So it does not matter how you structure things within an account. Because if that same conversion is being triggered across multiple campaigns, Google knows even those campaigns, that it’s not doing smart bidding, and it’s not automating, even in manual campaigns, it’s still pulling that data to make decisions for automated campaigns. Right. So structure should be driven by your business by how you need it to be to do what you want to do. That is the determining factor. Not that Google says Get rid of all your other campaigns. If you don’t need the other campaigns, great simplify stuff, right? It makes your life easier. But if you need those campaigns, for some reason, because you know that maybe you need different ad messaging in different locations. Fine, go for it. Google does not care. Alright, so I know I’m at time. I’ll be around for questions. But thanks for watching. And that was smart bidding and a very quick 25 minutes.
Unknown Speaker 24:36
Thanks for it enough. It’s gonna be a good hour long conversation, if not longer. So thank you, Fred. Wonderful. And with that, Fred, we’ll be back in just a few minutes to get into our q&a with everyone.
Transcribed by https://otter.ai