We’re at a point in PPC where automation is doing most of the work we used to do. There’s no escape from it, so we must find a way to coexist with the machines but stay in control of our PPC accounts.
Enter automation layering — you implement your own automation strategies over what Google or the other ad platforms are doing to put yourselves in charge.
And to tell you more about what automation layering is and how to go about it, we invited Joe Martinez, Michelle Morgan, and Aaron Levy — three of the top PPC experts and practitioners to our 50th episode of PPC Town Hall.
Here’s the full episode.
Full episode:
Transcript:
Fred’s introduction to the episode.
Fred: Hey, welcome to another episode of PPC Town Hall. This is episode number 50. So we’ve been doing this for quite a while and I’m so happy you’re watching this very special episode.
Today we’re going to talk about automation layering and automation layering is a very hot topic these days, right? I mean there’s so much automation coming out of places like Google, Microsoft, and the question is always in this world where everything seems to be automated, how do we as humans still remain relevant?
There’s this whole thing that we’d be talking about humans plus machines better than machines alone. But then automation layering came in because we started thinking about there’s quite a bit of work for humans to do manually when it comes to controlling and monitoring these machines.
Why don’t we try to automation layer and basically put our own automations on top of what Google’s doing so that we stay in control using sort of the methodology that we really like?
So that’s what we wanted to talk about today. “Robots fighting robots” is what some of the other panelists were saying. So I can’t wait to hear their take on this whole thing. But let me show you who we got for panelists today.
Let’s get rolling with PPC Town Hall.
Fred: Alright. Thank you for joining me, Aaron, Michelle. And who’s that guy in the top corner of the screen?
Aaron: I thought my headshot was outdated. Right.
Michelle: It’s just a Mr. T look alike. Don’t worry about it.
Joe: Yes, I pity all of you.
The panelists introduce themselves.
Fred: So yes, all of you have been on the show before. Thank you for coming back home. Now for those who may not know, you tell us a little bit about who you are. As my Alexa just turned around, she seems very interested in finding out more too.
So Joe, why don’t we start with you? Who are you?
Joe: I’m Joe Martinez, co-founder of the Paid Media Pros YouTube channel. Doing paid media for, I think January was my 10-year anniversary in the paid media space. So Hooray for me. Multiple anniversaries we’re celebrating this year.
Fred: Yes. And what about the other half of Pay Media Pros?
Michelle: Yeah, that’d be me.
Fred: So tell us a little bit more about you as well. Yeah.
Michelle: So my bio is going to sound pretty similar to Joe’s, except I have a different name and I look different.
So I am Michelle Morgan. I am the other co-founder of Paid Media Pros. Also been in the PPC space for about ten years now, working in agencies and in-house. All that good stuff. So excited to be here. Yeah.
Fred: And just when you said you look a little bit different, I was like, hey, anyone watching? If you’re good with Photoshop, we’d love to see it.
Michelle: I’d turn off my camera for the rest of the stream.
Fred: And then, unfortunately, we have Aaron hiding somewhere behind the bottom banner. There he is.
Aaron: First of all, Michelle, when you took a second there, I had that moment, like in elementary school when the teacher was calling you, but you weren’t paying attention.
Michelle: I was like, am I supposed to do my part? Yeah.
Aaron: Yeah. So I’m Aaron Levy. I’ve been in, I don’t want to say the number is now…I’ve been in SEM or Digital for 15 years. I run our paid search team at Tinuiti. We’re about 120 odd people, spend a lot of money.
And I’ve known Michelle and Joe forever and ever. So if we seem too comfortable with each other, that’s because we are. But honestly looking forward to talking about it because robots fighting robots just sounds like an awesome TV show.
Fred: I know, right? But these robots. Right. Let’s talk about what these robots really are and why we have our own robots.
So I want to start with what’s new with the engines and specifically, Google, I think is always most interesting. But anything there in the field of automation that you think we should talk about, Performance Max campaigns certainly come to mind for me.
So why don’t we start there?
Aaron: First of all, I’m a big golfer, as Michelle is. We’re about to find out shortly who is better than the other. I think it’s going to be neither one of us.
Michelle: Correct.
Aaron: First of all, this is going to be a swipe at Google. But in the golf world, there’s been a big trend of naming things like speed and Max.
The Max is always for the worst player that needs the most forgiveness. But no, I mean, Performance Max is obviously going to be huge. I think all of us are seeing it as the latest iteration of Google, saying, just give us a credit card, we’ll do it.
It’s going to be interesting to see because just based on our limited observations thus far, Performance Max does sort of what we expected to do that it just goes for the easiest conversion.
So deciding if it’s incrementally valuable or just taking credit for stuff that would happen anyway to tell. But it’s certainly a direction towards a more automated future and more input optimization rather than keywords or placements or topics like we’re used to.
Fred: Right. It’s a little bit the same problem that I think we had with smart shopping, where it kind of runs across channels, and it prioritizes remarketing because that’s the easiest and cheapest thing to capture.
Now with Performance Max, I mean, not everyone may be super familiar with it, but it basically runs your ads automatically across six sorts of channels that are part of Google. Right?
So it’ll do search, display, there’s YouTube, there’s Discover, there’s shopping, the other ones. Yeah. I think those are sort of the six main ones that they define. But really, where it’s headed is any place where Google can put an ad any place.
You just tell it a few things about yourself and your campaign, and then it runs with it and it’s kind of nice because in the past if you were sort of a newbie advertiser, probably not someone watching the show, but a newbie advertiser not very sophisticated, it was kind of a pain because you’d have to set up a shopping campaign and connect your merchant feed, and then you’d have to kind of give the same information for your DSA campaign and then kind of the same information for an ad campaign and then a whole bunch of campaign types that couldn’t really be automated.
So I can totally see the value for someone who’s a newbie making it quite easy. But then for people like us who actually care about the best performance possible because that’s how we get compensated by clients.
It is a bit painful, right, when everything becomes opaque and black boxish.
Michelle: Yes. It ends up being one of those things where, as you said, it’s basically you give a credit card, and then Google just takes the reins for you in some areas that can work really well.
I think a lot of it ends up being under the guise of we’re trying to help those small business advertisers who are doing it themselves and they don’t have anybody to help. So we’re simplifying things.
And basically, all it ends up being is that they’re taking things away or narrowing them and making it so it’s harder for the people who do have the budget and the traffic and all that kind of stuff to where it makes sense to invest more of the time.
I think Performance Max can be a good value add, but as Aaron said, it’s kind of hard to know sometimes if it’s just taking from campaigns you already have or if it’s actually adding value.
And it’s kind of tough to pinpoint when everything’s in the same environment.
Joe: There was one time that we just tested with a client and I’m working on that. They’re a startup, really. No brand awareness.
People aren’t looking for what they offer because they’re brand new. So it’s something where we’ve tried it just to get some sort of easier reach, instead of having to create since they do have a more limited budget, instead of having to create a separate display campaign, a separate discovery campaign, a separate YouTube campaign, just to have it all in one, just to get some baseline-type stuff.
But we purposely did that to keep the budget low just to see what it did and get some traction and hopefully get some learnings from it. And honestly, it did okay. There wasn’t any clear conclusion of, yeah, this is going to work great.
We’re going to keep building on this, but it gave some idea of, okay, we saw more engagement from the YouTube side. So now let’s go create a dedicated YouTube campaign and be a little bit more controlled with it.
What is the panelists’ take on Automation Layering?
Fred: That’s where I suppose automation layering comes in, Right? So you got this really complicated system from Google that we could not replicate and like you said, it drove okay results and you saw some success in some channels.
Now how do you take that success and automatically turn it into maybe a little bit more of a full-fledged campaign? Or how do you even figure out what portion of it is driving success and when that success has been established?
Now, for me, that’s automation layering, right? It’s like I don’t want to go into those campaigns on a daily basis and run the same analysis. I want that to be done for me. I want my system to tell me, hey, this automated thing from Google, it’s failing badly, so stop it or it’s working well, go and take a look at it and sort of spin it into something a little bit better.
What do you guys think about automation layering in these examples?
Aaron: I’m a little reticent, given that these Performance Max and some of the other new Google tools obviously have a bunch of data blind spots, which I think all of us know about.
Fred: Depending on your thoughts, let’s assume that not everyone listening knows about these databases and spots.
Aaron: Well. So from a Performance Max perspective, you get very little, primarily all that you can pull out right now.
And even this is a little hard. It’s pretty much impression reporting. And same for responsive search ads. So you won’t necessarily see as Joe peeled out himself, he figured out that YouTube was doing really well.
You won’t get active reporting on that. And there’s not necessarily a way to automate it. You have to understand and look into it. So when we think about automation layering number one, we want to prevent automation from being wrong, or at least being wrong in a place that’s destructive.
Like, if it’s wrong in a place that doesn’t hurt, that’s okay. If it’s wrong in a way where wrong causes problems, it’s a problem. So generally speaking, Fred, you make a good point about, you know, let’s automate this stuff.
Let’s say, like videos performing well, let’s fling it on YouTube and promote it, something like that. What we’ll generally do instead is try to think of a really funny cartoon character to relate this to, but it’s not working.
My robot brain is not as good. No, we’ll have an alert system like, hey, this thing was converted ten times this month. Like, you should look at it, maybe a desperate attention star robot.
But rather than have it make all these decisions where it could be making budgeting decisions or spending more money or things like that, that can make us a little sensitive. That and despite all of Google’s intelligence, their massive computing power or whatever, they don’t understand brand safety or what a brand is.
They can try it. But to be serious, it doesn’t think. So there are areas where we’ll have to think for it.
Joe: Like you said, the brand safety there, that part scares me a little bit. I haven’t worked on a finance client in a while, but the feed that I have years ago, every single piece of copy that you put in for banking and everything had to go through legal.
So it’s one of the things we’re like again, I don’t have anything now, but if you ever did again, I was like, we can’t have Google just spit out something or come up with whatever they want on their own.
There has to be some deep review for certain industries. So I understand why some people are scared or just flat out against it.
Aaron: Responsive search ads are very fun.
Fred: Those are fun. But at least you can pin right videos like what’s that going to look like.
Michelle: Yes. When it comes to all that kind of stuff, I think that the biggest thing that and really the only controls that we even get on some of those things are just making sure that we’re basically taking our robot and pointing it in the right direction to start with, one of the things for performance Max campaigns, when you set it up, you have to tell it what conversions you want.
And you have to be cognizant of the fact that some of those conversion actions can be created for you. It can be a phone call, it can be ‘get directions on the map’. I don’t know how many businesses actually make money anytime somebody clicks get directions on a map.
Now, that might be a lead into something that is useful, so it might make sense for you to try and optimize towards that. Or it might just be garbage, especially if you actually just want people to fill out your form online.
That’s not a good conversion action to even optimize for at all. So I mean, even just the first step in the set up process, it’s being simplified so that you don’t have quite as many levers that you have to pull and things to think about.
So it can be simplified. But if you’re even just accidentally pointing the robot in the wrong direction, it’s going to optimize for, like we said, the easiest conversion. And it’s probably going to be those get directions on the map when you’re trying to actually make online sales.
So even that first little input piece is going to be one of the bigger aspects of whether it’s going to work well or not.
Fred: Right. I would say that the cheapest conversion is cheap for a reason, right? I mean, there’s plenty of us bidding for the same keywords. If something is inexpensive, it’s probably because a more sophisticated advertiser has established that it’s not a high-quality click.
It’s not going to be a high-quality form fill. It’s not going to be a high-quality phone call. But if that’s what you’ve indicated that you care about to Google, I care about phone calls no matter what’s on the other side of the line, great.
We’ll give you those cheap phone calls and we’ll give you lots of them because nobody else wants to buy.
Joe: Yeah, because of it, I think I have shifted more towards the campaign level conversion tracking, which I’ll admit I didn’t do a ton when it first came out. We had the ability to do campaign levels.
But now to what Michelle said, there’s been times where just due to limited actions on websites just to get more signals, we have added just like click to call on your website as a conversion action.
Well, now it’s starting to lead to more junk and now we’re hearing like everything coming through is not that great. So like, okay, we’ll still leave it on there for some campaigns with lower touch points, but for certain ones we’ll just stick it to what is more important, which is the form submission ones.
Aaron: Well, and I think Joe and Michelle, you both raise an interesting point where Google is a closed system. Google is not your business. And so obviously those of us that have worked with our various Google reps, again, Google is trying to get as much information into the system as possible.
But that’s contingent on, number one, us being able to feed it everything that it really needs to make smart business decisions. And number two, making sure that and I’m going to be speaking about this at a conference, but making sure the human side of data is real.
It’s just an example that we saw from a real client of ours. They had this sales incentive thing going with their call center. So they got like extra bonus dollars if things got to be an SQL, if they sold something from an MQL to an SQL.
So guess what? The Friday at the end of that period, they had a huge spike in SQL because everybody wanted their $5 bonus per SQL. So then in turn, Google’s automated bidding, we feed it back and say, hey, we got all these SQLs, okay?
It doesn’t know that we had a promotion. It doesn’t know that these were like weird human motivations. So in turn, the system says I did great. So it’s going to do whatever it does really hard. But we have this sort of human leakage.
The challenge with robots fighting robots. Sometimes if a human feeds bad information to one of the robots, the fight is going to get dirty.
Michelle: It’s the whole idea of a system in and of itself. Right. The quality of data or input you get in is the same system you get out. I’ve used that for putting in things for lookalike audiences.
If you put in just some garbage audience of anybody who came to your website and it’s everybody in your CRM for the past 45 years, you’re probably not going to get a great thing coming out the other end.
But if you’ve got just your high paying customers, you’ll probably have a little bit better chance because there’s actually a pattern in there. So, yeah, kind of leaning into that, you’re only going to get as good a performance out of it as the indicators that you’re giving.
So going back to the idea of pointing it in the right direction, sometimes we think we’re doing such a good job and we’re the flawed portion of that relationship where we’re like, oh, I’m going to point a robot this way.
It’s like, oh, wait, that was right into the trash can. Maybe not the best choice there,
What do the panelists think about enhanced conversions?
Fred: Garbage in, garbage out right? Yeah, but that’s interesting because you’re even talking about something that’s already more sophisticated than what a lot of advertisers are doing, which is actually thinking about marketing qualified leads, sales qualified leads, things beyond just a form fill.
So I kind of want to shift here. There’s enhanced conversions now. Right. And I think a lot of the reasons why people haven’t done tracking of MQLs and SQLs is because it’s complicated.
You have to get the Gclid into your system. You have to persist it throughout your CRM. It’s not rocket science, but it’s also not necessarily super easy for a marketer who doesn’t necessarily have their engineering team support behind them.
But enhanced conversions, does anyone want to kind of explain what these are? And maybe does anyone want to volunteer?
Aaron: I was doing some reading about those the other day because we honestly haven’t used them very much. But my view of them is if you’re familiar with, like, a LiveRamp or an Epsilon or an audience enrichment platform, basically what it does is it tries to stitch data together more accurately than you could on your own.
And enhanced conversions can help with that. I might be talking about the wrong thing.
Fred: Yeah, you’re talking about the right thing.
Aaron: I read about a couple of things, but it starts stitching those things together.
Fred: Exactly. Stitching together. And so what’s fascinating to me is that instead of saying the Gclid is the thing that stitches together the original click. And Gclid, by the way, stands for Google Click ID. Right?
So the Google Click ID is unique when an impression of an ad is served, and then if somebody clicks on that impression, it has a parameter in it, a unique ID that Google can say later on.
It came from this search with this ad and this click. Now, the whole point as well, when the form gets filled or somebody calls you up, that Gclid feeds into your CRM system, and then two months later, after that customer becomes an MQL, an SQL, and finally a customer, you have that Gclid parameter that you can feed back to Google and say this new conversion, that’s actually more meaningful conversion was associated back to the Gclid.
And Google knows exactly what click it came from. So now it can go and machine learning can say, oh, well, these were the clicks that led to not just form fills, but actual sales. So let me prioritize more of these.
And to do that by the way you have to do tROAS bidding. You can’t do tCPA bidding. Right. And that’s a whole different story, but this Gclid is difficult.
And now here’s the thing that’s like so brilliantly simple from Google. When somebody fills out a lead Gen form, they probably put in their email address. When somebody becomes a customer, you probably know their email address.
So why don’t we use the email address to tie these two pieces of data back together? And assuming you’re not like getting thousands of clicks from the same email address, they can probably correlate that back to a single click.
And there you have it. It’s simpler offline conversion tracking that virtually anyone can do. You don’t have to have an engineering team anymore. And so that opens up this possibility of what you guys are talking about MQL SQL tracking to a lot more advertisers.
But kind of like, what could you do wrong in these scenarios? And I think you kind of covered it right, like human influences. But is there anything else people might mess up on?
Aaron: Well, we talk internally. And for the record, the other thing I was thinking about was Ads Data Hub, which is kind of similar but with mobile and video included.
But I think the thing that’s wrong and we talk about moving food around our plate and a lot of our conversations of Attribution or whatever it may be, it may be taking too much credit.
A person might have done a whole bunch of different stuff. Like, yeah, maybe they watched a YouTube video, but then maybe they downloaded a white paper somewhere offline or they read a newspaper article or whatever it may be.
And then Google’s, like, we touched it, we were there first, we’re involved. And so obviously you can make that deduction with frankly, any sort of conversion tracking, but that sort of automation saying we can stitch these together.
So therefore, like we did it, it can get a little bit aggressive. That said, the few that we’ve run, we’ve done it a few times, didn’t see a huge lift one way or the other. So that sort of conspiracy theory hasn’t really been proven out.
But of course, it still sticks in the back of my mind of “Are you just taking credit for stuff?”
Michelle: Because everything is always taking credit for stuff. Like stuff is not there. I mean, just the fact that even we have just a look back at windows, which a lot of the platforms obviously are much shorter than what they are now.
But if you have a bunch of people sign up for something and they first start in a CRM platform because they came through Google search or something like that, and then maybe an end of the month promotion, like maybe not on the sales side, but for the customer side.
So you send out something like the last two days of the month, we give 15% off, they come to the website, they convert and then it looks like, oh, our Google search went nuts. Not really.
I mean, it kind of did, but it kind of didn’t. So there’s a lot that is muddy in there, which you talk about here, presentations.
That’s actually what I’m talking about in my mind is trying to attribute performance to the top of the funnel. And the best takeaway is that it’s all just kind of a mess. So you guys don’t do the best you can. I think that’s the same thing with this is that there’s so many different things.
I mean, even just the idea that Fred, you had to tell us what enhanced conversion tracking was because it got slightly confused with something else. But also those are such bland words that like what does that even mean?
And we do this all the time every day. Like think about the people who don’t do this all the time every day trying to keep up with stuff. Like the idea of trying to get everything sorted out and making sure that everything is working well, especially when it’s just already this complicated with the automation just feels impossible to a lot of people, I bet.
Joe: So then how is.. This is an honest question, how is the offline data for these conversions being uploaded into Google? Who can do this? What type of CRM do you need and how do you get that back into Google?
So they can attribute everything just in general or for enhanced conversions?
Fred: For enhanced conversions, it’s the same as before, right. So in the past there’s a couple of mechanisms. You can either do a bulk upload or you can have an automated scheduled upload off of the spreadsheet.
Or you can use your CRM like you have a Salesforce integrator. But Salesforce is not a prerequisite, right? So anytime that you have an email address, you can basically feed that back and say this becomes this conversion action.
And so now you could have a conversion action that’s associated with an SQL, a different conversion action for an MQL, and a different one yet for your actual sale. And so on a daily, nightly basis, you feed it back in. Now there is a look back.
Joe: I was going to say I asked that for a reason, because Aaron and Michelle, have you ever worked with a client that had messy data within their CRM?
Michelle: Have I ever worked with a client who had clean data is the better question.
Joe: And that’s the thing too. That’s why I laugh at it. I know it’s like some bulk upload stuff too. It’s like there’s been somewhere if like salespeople or whoever’s managing the Salesforce team, they just don’t keep up with it.
But then if we want to try to work with it, that’s exactly how we could potentially steer the robots in the wrong direction. It’s just by uploading messy or out of date data because they forgot to upload customer status and everything like that.
What tools do the panelists use to create a layer of automation?
Fred: So it’s one of those types of messages. What do you guys think about that? Do you think sort of like that’s what you need to solve, that’s your value add?
Because the way that I’m thinking about it is like we used to do all these things like manage keywords, manage bids, like write ads, like all these little details in Google ads.
But now it’s more about how we steer the machine and this automation layering. So if their data in the CRM is messy, is that how we help them become better? Or do we sort of like say that here’s what we need to do.
Michelle: I do think that we end up being very much business coaches, more so than we used to be because it used to be something where we kind of play on our own little sandbox.
But now we do have to have all of that business data come in. I personally don’t really feel like having, especially with a lot of clients I work with, they end up being SaaS focused. They’re usually enterprise level, all that stuff which has its own unique challenges.
And one of them is the fact that it takes people a long time to go from a lead to an MQL and from an MQL to an SQL. On top of that, the data is messy and there’s only so much of a look back window.
And if you start uploading data to Google, that’s like six months old on a click. That was in June, it’s not going to be able to do much with it. So in that instance, I would probably maybe try and pass like an MQL into the system and use it as maybe a little bit of a guidance.
But we’re really talking about automation layering. I wouldn’t give the robots full control on this one. I wouldn’t import it and say, okay, and we want to optimize for these actions.
What I would do instead is basically just do a manual analysis and say, okay, on our side, we have all this data in Excel, and we know that this keyword typically has this return on investment.
So let’s make it so we can either make something dynamic or we create a different conversion action that has a different value associated with it.
So that on the conversion action, the initial one, you can have some sort of value associated and then maybe you optimize for Return On Ad Spend bidding or something like that, or you know, that something has to have a different cost per conversion because these leads are junk. These leads are good.
So these campaigns need to have a different target CPA bid model in place, even if it’s the same conversion action. Because otherwise I think it gets so bogged down trying to be too clever for itself, if you will, that it just automates itself into a circle of not being able to actually get what you want.
Fred: It’s a great point. Let’s shift a little bit. So what tools or technologies do you all like for helping you steer the machine and do automation layering? Yeah. Even like the reporting software where you like instead of moving the pieces around the plate, that’s to me an automation layer technology. Right?
What is a reporting suite that actually tells you the true picture?
Aaron: Yeah, we have a lot of toys. So number one, from an executional perspective, obviously using a tool like Optmyzr or Adalysis or things like that that can help build and dashboard a lot of things the way that we wouldn’t want Google to do and can help automate a lot of things that either we don’t trust Google to do or they’re not great at it.
Using those is helpful. We use those as well. We actually have a proprietary database tool in the data warehouse and Data Lake that we pump into Tableau for our own system called Mobius.
But having something like that and we have the luxury of having an entire team that can build these alerts and we can pipe in weather data and say like, “hey, it’s going to rain in the Midwest.”
So you’re going to have more bugs in three weeks. But having that information realistically, if any data sources in API, you can get it into your data warehouse.
How much you use that and fed to your point earlier, how many resources that takes and how many it has is something that you have to be a little bit judicious about. But I’d also echo from our free tool perspective, just using Data Studio is super powerful things plug in really easily.
You can set alerts pretty easily, even through Google’s own automated alert system and own rules like that. You can flash warning lights fairly easily and fairly free.
Just again, something that our philosophy does is not necessarily have the machine do it for us, but the machine tells us where to look.
Fred: Yeah. So an automated rule. I’m also obviously a big fan of scripts since we’re talking about unfortunately my session was not at the same time as Aaron’s. Now they are having us at the same time.
But Aaron is going to be really good. So go to his. I’m just talking about scripts turning ten years.
Aaron: Yeah. Fred is going to be brutal. Go to mine.
Joe: That’s pretty much what I’ve stuck to has been scripts and rules for the most part. I don’t have any sophisticated ways. It’s pretty much me for the most part. So I don’t have the huge data warehouses that Aaron’s company has.
Fred: Scripts, Joe, what do you do? You write them yourself? You got someone you know?
Joe: No, I pull from sources like you and Nils on Google or on Twitter as well. I rely on the smart coders and engineers to come up with those and then see if it fits. Not every… I’m not talking about just like the reporting ones, those automated ones, those.
Yeah, we can run in every single account, but some of them you’re going to have to look at depending on what type of account it is and the frequency and the volume. I do have a medical center in Florida in a small little radius.
So the data in there is pretty straightforward. Or we can pull information from Google Data Studio. We’ll look at the Attribution comparisons and then make some assumptions from there.
So we don’t need to get crazy. But on some of the bigger Lead-gen ones, that’s where we are. Even though I kind of ripped on some of the CRM data stuff, we’ll still import that information in and then run some more sophisticated bid scripts just based on understanding the lead times and everything of where we see them come in.
And those again, is based on volume and size of the account.
Fred: One thing that’s cool about scripts now is the new versions that are transitioning to Google Ads scripts instead of AdWords scripts, which basically just brings it in line with the new API that’s also currently rolling out, which is one of the reasons, for example, that WordStream is basically shutting down because they’re not transitioning to the new API.
But this new Ad script has the capability to set crowd targets, which the AdWords scripts didn’t have. Right? They kind of became very dated for modern PPC management and a smart bidding sort of world.
Now you can actually do this again. So I’m pretty excited about that capability.
Michelle: Yeah, I think that going back to kind of what you guys said. I think that having your reporting be automated is one of the things that is like the absolute best because one, we all hate reporting.
It takes forever to do anyway. But I would echo a smaller point that Joe made is that have your CRM data included in that and figure out how to enmesh the reports from the channels to your CRM.
Especially for a lot of the lead-gen accounts I work on, you’d be surprised how many people think that one campaign is doing great and then you will get the CRM data and it’s not.
Or they think that cost per lead is way too high, but then you incorporate it with the fact that people are coming through their customers. The lifetime value is huge, so it’s actually really not one lead pays for.
One customer pays for like a month or sometimes six months worth of marketing. So you need to have it all included to actually make sense with it. But then I personally really like automated rules, sometimes scripts, depending on what they’re trying to do.
Quite frankly, I do think they try to get too clever for what they’re trying to do. There’s a lot that goes into it and the more advanced you make it, that’s great. It just has a lot less use cases because there’s only so many places that you can use weather data because a lot of my accounts that doesn’t matter.
But I think that using automated rules is a great way to do it. It’s also just a little bit easier for people who don’t have quite as good of a handle on things. The automated rules are just a little bit more basic.
You just make sure that you’re not over optimizing on the same data, like don’t have your look back window be 30 days and then you run it every week because then you’re optimizing on that same one day like four times.
So make sure that you’ve got them spaced out appropriately. But I personally kind of like the pattern of going from manual monitoring of things like especially bids just for a basic example, and then noticing what the patterns are and then setting automated rule cadences so that I don’t have to do it, but then setting like calendar reminders to go in and say go check your automated rules.
Are they still doing what you want them to do? Are they still having the impact that you want them to have and just having that kind of check-in system of having the knowledge that we’ve got tapered into those automated rules and then checking back in every once in a while to make sure that the little machine we set up is actually doing what we wanted it to do? Yeah.
Aaron: I think Michelle and Joe, you both make an interesting point in a way of over automating. And Michelle like the phrase ‘over optimization’.
And Joe, to your point, if you were to pump weather data into a small regional facility, you’re going to make decisions that it’s stealing a cruise ship like a jet ski. You’re making too big of swings when you’re too small of a data set.
So it’s important as you go through all these things. And frankly, I see this when I look at some peers or friends’ accounts as you don’t necessarily have to use every tool at the same time.
So thinking about what tool will yield to seal our mutual friend Perna’s favorite phrase, what tool is the best route to return on time spent versus what tool is just doing a bunch of stuff for the sake of doing it.
I think it’s important to pay attention to that.
Joe: That’s what I think. That has really helped me embrace a lot of the automated bid strategies because do I really want to go in and look at bid adjustments from the audience level, then the device level, then the household income level, then the gender level?
To have a lot of that and do that for if you can trust that the data in there is right and your conversion actions are right, I’d rather have a script or an automated bid strategy.
Do that for me than having to do that every single week manually. Nobody wants to do that. So embrace that automation for sure.
Automation layering for bid management and messaging.
Fred: Let’s go deeper on that. Right. So let’s talk about automation layering for the sake of bid management. And we’ve already covered a number of things, but one thing that I’m looking at nowadays is value rules.
And so value rules is basically you get to communicate to Google for your conversions whether you want to value them more highly, depending on, for example, the geo or the device type or the audience segment.
So I could say if I get a conversion from California versus a conversion from New York, I’m going to value the one from California more highly.
And that’s kind of under the assumption that I haven’t done the sophisticated CRM integration. So I’m not really communicating my sale, but I’m maybe just communicating the form field.
And I think my form fills from California convert at a better rate. So this kind of is a way to do these bid adjustments on some of the factors that we used to do. But now you’ve got to think about, well, it’s not just what’s the likelihood of the person filling in the form, but the likelihood of when they fill in the form.
That being a better thing than another, I hope I’m explaining that sort of. Okay. So kind of curious, has anyone else looked at those bid adjustments? Because one thing that I love is the whole fact that you could look at your CRM and you could look at these factors that Google has previously looked at to predict conversion rate.
But you could look at it to predict through conversion rate without necessarily giving all that information to Google and just kind of steering the ship in little ways. Right. More budget towards California because that’s where my leads tend to become sales.
Michelle: Yeah, I really like those types of things. That’s kind of what I was alluding to, being able to create different values for different conversion actions, that sort of thing. This is just more sophisticated. So I really like this roll out personally, of being able to do this.
I haven’t used it yet, but that’s because I have too many areas where the data isn’t as sophisticated as I would like it to be for me to then tell Google the information. They would like me to tell them.
But I think that having this goes back to the same thing that I’ve been saying is like pointing your robot in the right direction. And these are just more controls that kind of help you lean into what is actually working.
So personally, I’m a big fan of these things. I think it’ll be really helpful to kind of layer those in with some of the automated stuff because you’re just giving it more data to go on, which is what all of the different types of automation pieces want.
Aaron: Really looking at it kind of like, I’m thinking a lot about my childhood and education today because they said how long I’ve been doing this job for. But I look at it like taking AP classes in high school so you don’t have to take the class in College.
You’re not paying tuition again for something that you already know in the sense that you’re not spending extra money for Google to learn that California is good. Now, obviously, if you don’t have the API integration, it might not learn that anyway.
But at the same time, again, this is assuming the API type stuff is set up. At the same time, you’ll probably get more out of relearning what you already know because what you already learned California leads are in the past.
People have been moving around a lot the past two years. So results that we had in the past or where people live or where their “base” might not be true anymore. So having that sort of persistent learning is nice.
But that said, Fred, to your point, there are some of these things where we know all this already. We don’t need to wait for Google or Microsoft or whoever to learn it. Again, much the same.
We haven’t used it a ton, but that’s partially because we have the, I call it luxury of working with larger accounts. So usually if a region is doing really better, we’ll just isolate it.
Fred: You guys have better data sets. If you can bring the Gclid data back into it and offline conversion tracking, that is better. But if you don’t have the luxury, then for sure, Value rules.
Joe: I haven’t used this yet, but they had a client that has warehousing data and they already know based on where the market is. I mean, a warehouse is not a virtual thing. It’s a physical place in a location.
So they know exactly where the main warehouse hubs are across the country. So they clearly see better conversion rates around where the main locations are. So that’s a perfect example of when you want to use it.
We have other clients where, yeah, they’re a national software brand. However, where their headquarters are converts way better just because that’s where they started. So they have much more brand recognition within the greater area of where their headquarters is based.
So that’s something where we can proactively say we know even for non branded keywords, we are going to do better in this area because we have better brand affinity in this area.
So we can be more aggressive here and always be aggressive and just feed Google the information. So there’s certain instances where, you know, it makes perfect sense.
Michelle: I think one additional piece that we have here is that we’re talking about all these performance-based stuff. But one thing that we haven’t talked about necessarily is like the human element, because there might not be a tangible reason why you like working with somebody who is from a specific state other than the fact that they’re easy to get along with.
There are also factors that don’t actually go into the numbers so that the machine can’t really learn it, but that you can guide certain things. Like if people in California are just nicer and you want to work with them more and there’s not like a monetary value for small business owners, that kind of stuff can be really important.
I have a number of clients who are like, okay, we’re going to target these areas but don’t target there because quite frankly we make money off of them, but I don’t like them.
So there are a lot of ways that you can have just like the human factor in some of these things that you can guide it to do whatever you want. It doesn’t always have to be numbers and data driven focus.
You can also make business decisions for yourself with these different controls too.
Aaron: Well, number one, this is funny being a person in California, a person in Philadelphia talking to two people in the Midwest, talking about people we want to work with who are nicer.
I don’t think anybody wants to work with me, but I think the other factor on that front too, looking outside of just outside of just B2B, SAAS, whatever it may be, is shipping costs, freight costs, supply chain maybe.
I think back when I did actual work and had a couple of clients, we had one client that could guarantee one day shipping anywhere that was within an hour of I95 on the East Coast.
So we had literally campaigns that targeted a road so we could say one day shipping conversion rates would be way higher on this little Causeway, it would be two day shipping and say this because he’s not here, sorry, Kirk Williams, we excluded Montana because it was too expensive to get stuff there.
So if they would run free shipping campaigns, they would lose money to stuff to Montana, Wyoming, the Dakotas, stuff like that. And so having a tool like this where we could say, look, we’ll ship orders to Montana, but they have to be at a way higher margin than we would get in a different place.
Okay, let’s have a much higher rise target there. Then along our dear friend I95, we could be just fine and go cheap.
Fred: Right. And I think what’s interesting here is that you have these sorts of decisions based off of stationary things like roads and States. So you make a one time decision to point the machine in the right direction.
But then when it gets fascinating is like you said, while there’s supply chain issues. So maybe it used to be possible to do the one day delivery on I95, but now half of my truckers are out sick with COVID, so maybe I can’t deliver on that.
So for a very short period of time, I need to modify my targets to account for that and I think that’s really where automation layering shines. Right? I don’t want to have to go in and modify 100 campaigns across a bunch of accounts.
I just want to have a thing where I put it on a spreadsheet. I’m like, “Hey, delivery time from one day to two days. It knows what campaigns to go and change, and it just handles it for me.”
And then when the truckers all come back in and they can do the delivery, go and change it back. All right, but let’s talk a little bit about automation layering for targeting.
And we could also get into messaging. Right? So targeting, placements, keywords, all that good stuff, and then messaging. RSAs, have you seen any good examples of automation layering in those fields? No.
Aaron: RSA’s data is still sort of limited or tools aren’t very limited. So you can’t really layer much. We’ve tried, believe us, we’ve tried. You get a lot more success for regional campaigns if you put the name of the neighborhood in the ad.
I don’t want to build campaigns for every neighborhood in the world. But so I’ve tried to do that with business data feeds, and our sales aren’t really there yet. What I’ll say from a targeting perspective, again, stealing other people.
But audiences has been super successful for us, and I’ll reluctantly call the new Broad Match automation debatable.
Fred: But hey, well, aren’t you the one who said keywords are dead? So to even, like, acknowledge that Broad Match exists?
Aaron: Did I say keywords? I didn’t say keywords. You did.
Fred: Okay.
Aaron: I said broad match. But no, I mean, you sort of set us up there what we’ve done in some cases, like example, that we actually want to use a search award for.
We have this company that sends out a ton of direct mail. So, yeah, we’ll have a, “regular campaign” that’s targeted based on that word that I just said. I wouldn’t say with tCPA, but then we would also have a campaign that was basically automated upload of when people got mail.
So, like, when a mailing list went out, their direct mail software would send in and be like, “Hey, here’s the people that got mail.” We bid through the roof on those folks if they search for anything that was even remotely close.
And it worked great because we knew that we were reaching the right people. We knew that the brand was already in the right mindset. And so in that particular case, words mattered but didn’t matter.
It was a little bit more of a thing of where we have this information that we know that they’ve already been contacted by us. We can do it easily enough.
We don’t have to upload the list by ourselves every time, and we can kind of get rid of it after a couple of weeks because we know that the postcards have already been dropped, recycled, and probably not opened.
Michelle: Yeah, I think the same thing can kind of be done for dynamic search ads as well. Like, you don’t have to necessarily use broad match. You can also use your website, which I think sometimes helps if your website is a good fit for that.
So if you want to decide if your website is a good fit, just reverse engineer the keyword planner. Go put your URL in it and see what keywords it suggests. If they’re good, you should use your website for DSA.
If they’re a bunch of garbage keywords about contact us and your Privacy terms. Probably don’t use your website for DSA, or you use page feed so you can control which pages it’s actually pulling from.
But yeah, going back to the idea of Google’s new favorite, the old broad match keywords with smart bidding, which are going to get you those low hanging conversions, that sort of thing.
The only real time that I like doing that is if I’ve also got a remarketing audience layered in. And Joe and I talked about that at a SMX presentation that we gave not too terribly long ago.
It’s like if you can control for the fact that you know the audience is the right fit, I’m a lot more lenient on what search terms people are typing in and misspellings and all things that are tangentially related and we can lean into the automated bidding.
But if I don’t know exactly who the audience is, then I’m a little bit more weary about doing that sort of thing. So it gets a little fuzzy. I really do like utilizing audiences and then kind of letting the machines take over within that confined playground, if you will.
Joe: We like using Retargeting with DSA too. That just kind of opened the floodgates up a little bit, because with DSA you can see search terms and certain headlines because we don’t create the headlines for DSA and that can feed you new ideas of what to potentially test for your responsive or if you still want to do expanded text ads just to get more information.
But you ran that on a DSA remarketing campaign, whether you use a page feed or your entire website.
Aaron: I think if one other thing to say, this is going to go maybe a little bit outside of PPC, there’s other marketing things that we can do. But if we run discovery ads, if we run Disco, if we run RSAs, if we run Retargeting, set an alert for audience weirdness, or have it get audience insights sent your way and then shift that over to your content team.
The best example that I can give get pointed to personal experience, but we had a wedding dress client. What do you think thematically people were going to in the remarketing audience?
Michelle: Dieting.
Aaron: Yes. Where else were they going thing I was guessing there’s that there’s obviously like Pinterest. There’s a lot of people looking at stuff index really high for weather.
Michelle: Yeah. There you go.
Aaron: So, of course, like, okay, now you have a content strategy for okay, what to do if it rains on your wedding day. How to aim your ceremony so that your audience isn’t blinded and you don’t have to wear sunglasses with your dress, whatever it may be.
But finding those things that are a little bit weird, that are highly indexed. And Fred, I saw you laughing. That was a real scenario. I performed at my sister’s wedding and could not read my script.
So literally all that you see in the photos is me looking down and my lovely little ball spot shining in the sun. Yeah, close to the heart. Anyway, but using things like that, using, again, all the wealth of information that we still do have within search to spike up other stuff that you can see to other aspects of the business.
Number one, it’s a good way to get other people in the business to care about search. And number two, it’s a way to increase the value of it. As CPCs get higher and higher every year, it’s a way to gain additional value outside of just they bought wedding dresses.
Joe: There’s one thing that’s kind of related to this. For any of the awareness type campaigns that you may be running on Google, it’s just there are auto created or auto-suggested audiences within Google.
When you’re looking at your audience targeting options where you can say it’s not really an in market audience, but it’s based upon your current keywords, maybe you want to try this audience targeting for your top of funnel campaigns. And some of those we found have worked better than some of the custom segments that we’ve created.
So it’s just something that we thought we tested out. But it’s all recommended stuff just either based on website behavior and all that type of stuff that you’ve done with your previous campaign performance.
Google has auto-created some additional targeting options for you to try.
Wrapping up
Fred: Great. Well, I love all the examples that you’ve shared and to talk about automation layering, both from a big company Tinuiti, with lots of PPC people, big budgets, sophisticated tools, all the way to Joe and Michelle kind of picking scripts from where they can find them using different automation.
So thank you so much for sharing all these examples. This has been fantastic. Also a lot of references to conferences. So SMX Next. I believe that’s where Michelle and Joe did a great session on automation layering that is recorded.
I think you do have to pay, but if you want to see that that is out there, look for SMX Next 2021 and then all of us are going to be speaking at HeroConf Austin, assuming that travel restrictions don’t all of a sudden pop up again.
But we’re all very hopeful to be there next week. By the time you see this, probably that session will have happened. But talk to us. We’ll share those slides, we’ll try to make those videos available.
So anything else? I’ll give each of you a quick minute to maybe tell people where to find you, what you want them to do. I would love people to actually go and buy my new book, ‘Unlevel the Playing Field’. It came out January 26.
So go and buy a copy of this. Talk about a lot of the concepts we cover today. Aaron, let’s start with you.
Aaron: You can take two approaches to automation. In my mind, you can either defend or you could befriend. And so I would encourage people, rather than trying to preserve what you used to do and try to have old search last forever.
Don’t defend against it. Just befriend. Understand what the robots are trying to do, what makes the robot happy, what makes him do his little robot thing or her. And then, of course, you can learn the pitfalls.
And so rather than trying to stop it from doing what it’s good at, instead, let it have a little freedom. Let it learn, and just make sure that you don’t let it fall off a cliff.
Fred: Friendly robot. Good advice. Michelle? What about you?
Michelle: Yeah, I would say I don’t have clever little words that rhyme and that kind of thing like Aaron does. I haven’t thought about that much, but I do think that that’s the right approach to it.
So set aside some amount for testing some of the automation pieces. You’d be surprised by some of the stuff that works better than you think. And you’d be surprised by some of the stuff that works absolutely atrociously.
And you thought that it might be the right way to go about it. So test some stuff. Don’t hold so tightly to what you used to do, especially if you’ve been in this industry for such a long time.
Like, it doesn’t help to complain about it. It really doesn’t. But don’t abandon those principles. Right. We got here for a certain thing. So try and think, maybe take a step back and say, okay, I know this is a foundational theory of the way that I always have done stuff.
How can I apply it here and utilize that to move forward and make sure that I’m staying up with the times? Because I can guarantee your competitors are probably trying to hold pretty tight to things.
So if you are staying up with the times and they’re not, it’s a great way to beat your competitors.
Fred: That’s how you ‘Unlevel the Playing Field’.
Michelle: Right, right. Buy Fred’s Book.
Joe: We need an affiliate link to that book.
Michelle: I know.
Fred: Listen, everyone likes these ideas, and these books come from conversations with really smart people like everyone on the call today. So really, a lot of credit goes to them as well.
Michelle: I’ll just take a kickback instead of an affiliate link. Thanks so much. Go ahead, Joe.
Aaron: So I’m an author. That’s awesome.
Joe: I think Aaron and Michelle kind of said what I would typically recommend. You’re going to find the complaining crowd on Twitter that likes to complain that things are changing, but it’s always going to change.
There’s always going to be new features, technology is always going to advance, user behavior is always going to change so embrace it as soon as possible. See where it figures in with your account.
You don’t have to blindly implement everything that Google lists in the recommendation but at the same time don’t ignore them or decline them because sometimes those recommendations or the automation can be very beneficial for your account.
There’s no one size fits all. You gotta do what’s better for your business but always make sure that you’re keeping your business goals first over Google.
Fred: Great. All right, so look up Joe and Michelle at Paid Media Pros, find Aaron at Tinuiti. We’re all on Twitter, we’re all at conferences so try to keep in touch. We love hearing from you.
Thank you for watching this episode. If you’ve enjoyed it and want to see more, subscribe to our YouTube channel and we’ll be back with another episode very soon. Thanks for watching.
Connect with the panelists.
Frederick Vallaeys: Twitter | LinkedIn | Unlevel The Playing Field
Aaron Levy: Twitter
Joe Martinez: Twitter | YouTube
Michelle Morgan: Twitter | YouTube
Takeaways
- Automation layering is how you shape what ad platforms do to suit your brands and accounts.
- Look at automation as a complement to your efforts rather than as a competitor.
