Artificial intelligence is radically altering the digital marketing landscape. Today, we’re talking to a former Google AdWords evangelist, Frederick Vallaeys, and author of Digital Marketing in an Artificial Intelligence World.
Frederick is going to tell us a little bit about what’s coming next, especially in the PPC, which is pay per click, avenue. I know sometimes we think of AI as robots taking over our jobs and a Jetsons type of reality, but Frederick tells us that’s not really how it works. With his expert guidance, you’ll not only survive but you’ll also thrive in tomorrow’s AI world.
Frederick Vallaeys: I’ve been blogging for Search Engine Land for a really long time, and probably about a year and a half ago I was thinking of some topics. I’d just been reading a number of books on this and on the topic of automation and artificial intelligence, reading in the newspaper how everybody’s concerned about their jobs being taken over by the robots. So I started thinking about how all of that applied to the space that I work in, digital marketing.
I started writing a couple of blog posts on the topic and really had a lot of readership. They were some of the most popular posts on that blog for that year.
So I figured, yeah, I’ve hit on a topic that people really care about that people are either passionate about or concerned about. So there was something there, and then I kept that as a theme throughout some of the other posts that I did, and eventually I had written so much in blogs.
I was like, this is actually a pretty juicy topic and something we could turn into a book. It’s probably worth bringing it all together in one place so that people don’t have to find blog posts and read them in the right order. So let’s distill it down into one solid chunk and give people one thing to read and be up on what’s happening in space.
The Robot Threat
Rae Williams: Did you find a lot of people genuinely worried about, you know, I wouldn’t say robots necessarily but automation taking over their jobs?
Frederick Vallaeys: I think there’s sort of two classes of people. One class doesn’t want to talk about tools and automation, and they see it as a true competitor to what they do. It’s not just artificial intelligence. In that case it’s just a tool that helps you calculate the right bids through a spreadsheet because they think, “Well that’s what I’ve been hired by the agency to do. That’s what I’ve been hired by the company to do,” from a marketing perspective.
So they are worried about any technology that makes them more efficient. They don’t see it as an efficiency tool and something that allows them to go and explore new portions of the area that they work in, but they see it as a threat.
Then there’s another class of people who are actually excited about this, right? They want to do things that make them more efficient, but then they have questions—how does the technology really work and to what degree can I rely on it? So for them to better understand the technology that’s actually doing this is really helpful, because now it gives them perspective on how they fit in that bigger picture.
So what is the machine going to do and where am I, as the human, still complementary, still needed? How can I work together with the machine to actually make the results better as opposed to just handing off everything to the machine and then hoping that things work out really well.
Future Proofing Us All
Rae Williams: So when you say future proofing, are we talking about future proofing in the sense of protecting your jobs from AI or making sure that wherever this technology is moving, your company can move with it?
Frederick Vallaeys: Well I think it’s both, right? And it’s not a protectionist view, it’s more figuring out where your job is going to evolve to. So maybe it’s a little bit of the latter or more of the latter.
Because I don’t think people are going to be doing the exact same roles five years from now that they’re doing today, right? They may still work in the same field. They may still work in digital marketing, but we’re not necessarily going to be manually doing the things we’re doing manually today.
And that’s just because some of the technology is not quite there yet to do things like automatically generating ad text. That still requires a pretty heavy human component and understanding the psychology of marketing and how people will respond to the messages that you put out there.
But who knows? Machine learning is evolving so quickly, and all of a sudden it may be really good. So that it’s more about understanding the what—are the machines getting better at, what’s driving that? And so how do I put myself in a position about I’m going to have something really valuable to sell to my customer.
And so either that customer is you as an agency selling to companies for whom you market or your customer could be your boss if you work in house on a marketing team, right? So how do you make sure that you’re still providing as much value as possible, but not necessarily positioning yourself as someone who’s just doing the same thing that a machine could have done?
In which case, you’re probably more expensive. It’s going to be more expensive to pay your salary than to pay a machine to do the same thing, and the machine probably will do it faster, more reliably.
But that doesn’t mean there nothing left for you to do. It just means you can be more strategic, take on entirely new roles within digital marketing that you may not have had time to do in the past because you were so busy doing the stuff that should have been automated but maybe wasn’t automateable at the time.
The Future of Automation
Rae Williams: How do you envision us coexisting with technology and with automations that are kind of doing a better job at some of the things that we could do or are less expensive? Where do we fit in as humans with our skills in doing that?
Frederick Vallaeys: Yeah, so well there’s a couple of roles. So one of them is that we as humans still have empathy, so we are still working with a client, a stakeholder and so the technology is not always going to be perfect, right? There’s going to be mistakes made. There’s going to be the communication that we have to have around what are your goals as the advertiser and then we can figure out as the practitioner, what technology do I put in place to help you achieve those goals? Sometimes these are hard conversations that still require the human touch.
I was reading an article in the newspaper yesterday of a hospital, and they had one of these Segways that has a little video screen on top of it so that the doctor can virtually walk around the hospital. One of the Segways comes into a room and basically tells the patient that they had a few days left to live, that they were not going to make it.
And the family of that patient was very upset, because they were like, “Well, that’s kind of a situation where we would like to have been able to talk to a human in person.” And sort of have that empathy from the doctor as opposed to kind of talking to a machine.
I think it’s the same in digital marketing, right? You still want to have that human touch and kind of communicate that understanding. So that’s one role.
The second role is very much oversight. So the technology that we see in PPC and automation today, they tend to be point solutions. So you may have a solution that does a really great job at managing bids and a different one that does a really good job of finding new keywords for which to show your ads. And you can have another technology that handles the budgeting. So making sure you don’t spend more than you want it to spend for the month, but they’re all separate solutions.
Putting them together in a way that they don’t interfere with each other and monitoring that are individually doing a good job and working towards the bigger goal that you have, then that’s a second role.
I called that the pilot of PPC. If you think about an airline pilot, a lot of the systems in airliners are automated, but you still need that pilot to oversee what happens and someone who’s really good at taking action.
If the machine fails in some way, didn’t put the pieces together in the right way, the human pilot is still there to course correct and make sure things don’t end up really badly. Same thing in PPC. So you’ve got those roles and then there’s also kind of a competitive pilot situation. So if you think about PPC, it’s a very competitive space.
It’s actually really interesting because oftentimes when you think about automation in the world today and the things that we’re not going to have to do 10 years from now, driving is sort of at the top of that list with all the self-driving technologies coming out.
Now, what’s interesting about self-driving cars is the whole community benefits if you have self-driving cars. Yes, it may take us two minutes longer to get to work, but we didn’t have to spend the time paying attention to the road, right? So that’s actually two minutes longer, but it’s also 22 minutes saved from a tedious task that none of us really want to do, sitting in traffic, that we can now do something else with.
Now if you think about PPC, it’s not collaborative, it’s actually competitive. If somebody else is using an automation technology and I figure out how that technology is working and what decisions it is going to make, I can actually try to get a competitive edge. If I know it’s going to make a very conservative bid adjustment, I can make a more aggressive bid adjustment and maybe beat my competitor and take over position one on Google because of that. So that’s the elder role.
So even if we have all this automation, we still need humans, and humans will still want to do this. They will want to look at how can I beat the system, how can I get an edge for my company so that my boss is going to be happier with the performance that I drove.
Rae Williams: So besides PPC and just kind of the advertising sense of things, where else do you see technical automations and things like that in digital marketing?
Frederick Vallaeys: Yeah, so when I think of digital marketing it spans a number of things, right? So their social media marketing, there is SEO and there’s Google display network and I think all of these systems are getting a level of automation added to it. Some of them have already been doing it for a very long time. So if you think about the Google display network and display advertising, Google has long had these ways of figuring out websites that might be relevant to what your audience cares about and then showing the right ads on those pages.
That would have been completely impossible to do manually, but some machines have taken it on. Another good example there is think about YouTube videos. So matching ads to the right YouTube videos, I mean there’s so much content being generated, there is no way for humans to classify that. So the machine will classify it and figure out, “Oh, this is a video about a makeup tutorial. So maybe we should show ads for Maybelline. That seems to be relevant to what’s happening in this video.”
All of that’s automated. That’s machine learning, that’s artificial intelligence and so those areas are going to continue to evolve and it’ll just continue to do a better and better job of matching the right ad to the right customer, regardless of whether they are interacting with a news website, a video, or a search results page on Google.
Working with Technology
Rae Williams: If you had to choose just one unique idea or story from your book and just from your expertise in general, because there are a lot of concepts here, what would that be and what is that one idea that a listener or a reader of your book can take action on this week?
Frederick Vallaeys: Yeah. Well, so one big idea and one big story from the book is that Garry Kasparov was beaten by deep blue chess machine back in the 1990s. So a long time ago, and this was a big deal in the media, like a computer is better at playing chess than a human.
But Garry Kasparov did some interesting research and he said, “Well actually, if you give a decent human chess player a semi decent machine, like not a supercomputer necessarily, but that human has a great process for working with that machine, that computer, the two of them will actually beat the super computer any day of the week.”
And I think it’s the same in PPC. And that’s sort of the takeaway here, is you’ve got to know how to work with the technology that’s out there and to work with it the right way. You really have to think about your process. So I think in PPC there’s still a lot of ad hoc tasks being done.
I think the digital advertisers who are going to be the winners over the next decade are the ones who figure out what is the process, what’s the workflow, right?
How do we put the pieces together—great humans plus great machines and great technology. What’s the order of things and what’s the sequence that we do things? And it kind of equates to, If you think about the most successful coffee chain in the world, Starbucks. Do they have the best Baristas? They probably have some amazing Baristas, but they’ve got 100,000 of them, right? So are they all the best ones in the world? Well, that’s kind of unlikely.
But what makes them great is the fact that they have an amazing process. So they have good people who know how to follow the process, they have great technology in their coffee machines that they have and they’ve put those two together and that’s how they became a winning coffee chain. And I think we haven’t really seen that into PPC world.
There’s not one agency that stands out as the clear winner but I think if somebody figures out what’s the right process, what’s the right technology to put into that process, I think that’s going to create some really interesting opportunities for advertisers.
The Price of Ignoring Progress
Rae Williams: What happens if we don’t do anything with this information? If we’re not embracing automation within digital marketing and at large?
Frederick Vallaeys: Yeah, so if you’re not the one who takes action on this, somebody else will, right? And the whole point about the technology and the machine learning and the artificial intelligence is that we’re at this point in time when we’re talking about this and why are we talking about it? Well it’s because it’s actually become really, really good. The reason it’s become really, really good is because of Moore’s Law—that computing speed doubles roughly every 18 months.
Artificial intelligence has been around since the 1950s, so why are we talking about it in 2019?
Well we’re talking about it because now every 18 months when you have one of these doublings, it is such a meaningful leap forward compared to the tiny leaps that were being taken in the earlier days of artificial intelligence, and that it’s kind of mind blowing, the kind of stuff that the machines are going to be able to do 18 months from now.
We have sort of an inkling of what they’ll be able to do. But kind of the degree to which they’ll be good at doing bid management and finding new customers for us, it just going to be super meaningful. If you just sit on the sidelines and ignore this technology because you think, “Hey, you know, maybe it’s good but it’s not good enough, I can still do a better job manually.” Well that may very much not be the case within the next 18 months. So you have to be testing this.
You have to figure out where do you put it into your process, and you have to figure out what are you as a human going to do that’s complimentary to all of this technology that’s out there. Only then are you going to be successful. If you sit on the sidelines, somebody else is going to figure this out. You’re going to have a really hard time selling your services to anyone.
Rae Williams: Give us a few examples of your favorite success stories or how some of the kind of lessons and teachings in your book have affected other people’s lives.
Frederick Vallaeys: So when it comes to success stories, the one angle to look at is bid management. So Google has done a tremendous amount of work in automating bids through what they call smart bidding and if you even look six, seven years ago, like the technology was adequate to not have you have to do the translation from a cost per acquisition target to a cost per click bid.
But it wasn’t all that good, right? It was looking at a lot of signals, but machine learning wasn’t that great. So results were mediocre at best and humans could often still do a better job but then those agencies that are have actually kept investing in this and kept figuring out, “Well, how do we give the machine more data about our business and the needle that moves our business?”
They’re actually doing quite well, right? The machines that Google have are the best in the world at figuring out what is the right bid adjustment to set based on the geographic location of the user or based on the time of day or the type of device that we’re using.
But those advertisers who’ve been able to bring in something that’s unique to their business, like how much snow fell yesterday, that’s generally not that useful to the majority of advertisers. But if you happen to have a ski resort and you’re selling lift tickets, that’s a really meaningful thing. And Google doesn’t necessarily look at that because the stuff they analyze tends to be more on a global level, right?
So time of day, geography, that applies to all advertisers. But those advertisers that have figured out how do we take what matters to our business and feed that into the Google machine learning system by changing the target CPA, the target row as, based on what we know is going to happen to our business—those have done really, really well.
Myths about AI
Rae Williams: What are some of those fictitious ideas that people have or myths about digital technology that need to be debunked?
Frederick Vallaeys: Yeah, the media likes to talk about artificial intelligence as like this humanoid who’s basically going to sit in your chair and do your job for you. I think a lot of people probably know that that is not exactly what’s going to happen, but it’s like this almost sexy vision of the future.
When you look at what’s really driving automation in digital marketing, it’s very much about machine learning, and machine learning is a lot of statistics and math and it’s actually very dry and quite boring in a way.
And there’s a couple of points in that. First, if you as an advertiser get to figure out one point solution that you can improve upon that could be really meaningful to your business.
You can actually build your own machine learning technology using some of the open platforms that exist from Google, from Microsoft, from Amazon. I think these are really exciting because like I said before, you can take what you think is a meaningful factor that determines how much you sell, how much profit you make.
And you can feed that into the machine and now you can start correlating that with things like what’s happening in the stock market, what’s happening with the weather, what’s happening with other things that that may be meaningful to your business.
So say that you are a hotel company, do you think you sell more hotel rooms weeks around when you have a big conference in town? Probably, right? But to what degree does that impact it and which conference impacts and how much?
These are great questions that people oftentimes don’t have the answers. They know it’s meaningful, but now you could actually deploy machine learning to figure that out. Once you know how much certain events impact your business, you can take that back and you can start to update bids into system, or we can change the ad text to really talk about this event that’s happening and maybe drive more sales to have it.
What Is Machine Learning
Rae Williams: What, in your opinion, is the one thing that people need to know the most? What do you think is the least known thing as it comes to digital marketing and technology and automated technology, artificial intelligence that people need to know?
Frederick Vallaeys: Well one thing that I like people to think about and know is that when we talk about automation and artificial intelligence and digital marketing, it’s often driven by machine learning. That name says, machine learning, it’s a learning system.
So you can’t expect it to be instantly perfect. You have to teach it something. So first of all, you as the human having done PPC and digital marketing for a long time, there’s probably something you can teach the machine.
You can give it examples of what you’ve done in the past and the machine can learn from that and then start to automate some of those same decisions for future campaigns that you run.
So that’s one thing, and then the second thing is that it is a learning system, right? So if you hire a new human to do some tasks in your digital marketing, that human is going to make mistakes, and that’s natural. People make mistakes as they learn, that’s part of the process of learning and it’s the same for machines. They’re not always going to be perfect.
But humans have this funny reaction where if another human makes a mistake, they talked to that human and they were like, “Well, you’ve learned, I’m not going to fire you. Try it again but do better next time.” And that’s fine, and that’s how people develop into rock stars. But when it comes to technology, that machine makes a mistake and the human feels, “Well there’s nothing I can tell this machine to make it do better.” So they turn off the technology.
They turn off the automation and then that puts you on the bucket of people sitting on the sidelines who may fail because others are doing a much better job at this. Really what you should have thought about was the machine is also learning.
So how do we give it better information so it can make better decisions and get smarter rather than just turning it off and figuring, “Hey, this thing doesn’t work.” It actually does work, but it needs time to learn.
Behind Digital Marketing
Rae Williams: Give me a little bit of background about you, just kind of how you got into this journey, how you learned some of the things that you’ve learned.
Frederick Vallaeys: So my background is I spent 10 years working at Google on Google AdWords, nowadays called Google Ads. I came in because they needed someone who spoke Dutch. So I ended up translating Google AdWords into the Dutch language and doing a lot of the customer support in the beginning. But then we found that a lot of people just wanted to know more about how the whole system worked and best practices.
So I created a role of a Google AdWords Evangelist, and I started going to conferences and customer meetings and basically just teaching how you do a better job with this PPC and digital marketing. It was a ton of fun.
I got a lot of feedback in these events that I did, in these meetings, and I took it back to the product teams so that Google would build better solutions for its advertisers.
One of the teams I was on was the quality score team, and that was really the original machine learning system within the Google ads system.
It was basically figuring out or predicting what kind of click through rates you could expect for certain keywords in certain searches. So it was using a huge machine learning model to figure this out, and that kept evolving. So that was a lot of fun working on that.
Then eventually I decided to leave Google and start managing a few accounts on my own, but then I was like, “Oh my God, this is actually way harder and way more manual than it should be.”
I couldn’t grow my practice. I couldn’t grow my own agency because I spent so much time talking to people. I have very little time left to actually do all of this manual, tedious work that needed to be done. So I started building tools and technology to put myself in a position to be able to actually deliver on the work.
And then I figured, “Well this is actually pretty cool technology. Maybe other people might want to use this.” And that’s when Optmyzr started. So me and my two co-founders, we started a company to make PPC experts more efficient, more efficient with their time, but also better ad driving.
And then me and my two co-founders, we started the company around us because we thought technology that we were building would be really useful to a lot of advertisers out there. So we would make them more efficient when managing their campaigns, specifically efficient with their time, but also more efficient in terms of giving them insights that would lead to better optimizations.
We have these one click optimizations. So that someone can literally go to a page with suggestions, and these suggestions are based on machine learning and algorithms that we run. And with a single click you can accept them or you can layer on top your own insights.
So that’s my expertise. I worked with tons and tons of customers, advertisers, talking to them on a daily basis, hearing what pain points they have and helping solve those. But also still have a really close relationship with Google so that I understand what’s coming next and hopefully help advertisers prepare for that.
Connect with Frederick Vallaeys
Rae Williams: Is there anything else that you want us to know or that you want to kind of leave with people about your industries, digital marketing, or just intelligence in general?
Frederick Vallaeys: So one thing I hope that people take away from the book is that there is very much a role for humans to play, even if it looks like Google is automating everything in digital marketing. Humans plus machines plus a great process that’s going to give you the best results.
But you have to be ready to adapt and change.
Now, Google Ads has always been a very fast moving product with lots and lots of changes happening every single year. So if you work in this space, you’re probably used to changing how you do business. Just continue that, things are going to change.
There’s going to be more automation, and expect that a couple of years from now what you do as a digital marketing expert may be completely different from what you do today. But you’re still going to be super valuable in this space and you’re going to drive a lot of value for your customers, if you learn how to work with all of this artificial intelligence.
Rae Williams: How can your readers and our listeners contact you?
Frederick Vallaeys: Yeah, so if they want to get in touch, I’m on Twitter @siliconvallaeys and they can of course also contact me through my company Optmyzr.
So if you drop a line to the support team and ask for me, they’ll put you in touch with me. That’s at www.optmyzr.com.