In this episode:
01:20 – Authority and awards
05:25 – Machines beating humans
08:52 – Starting at the bottom: machines bidding
12:07 – Games and presentations
16:47 – Next layer: targeting
18:23 – AI gone rogue
20:07 – The human element
23:56 – At the top: messaging
26:30 – What technology is learning about us
27:49 – Can AI write your ads?
30:20 – The exponential curve
33:46 – From now till April
34:42 – Suggested focus for 2019
37:24 – Should you be doing this yourself?
James: Welcome back to SuperFastBusiness.com. This is Episode 607. And I brought back a very popular guest, Mike Rhodes. Welcome.
Mike: Thanks, mate. Good to be back.
James: This is not your first rodeo on this podcast.
Mike: We have known each other something like 12 years. So yes, it’s been a history there. It’s just great to be here. Thank you for having me back on.
Authority and awards
James: I remember when I saw you presenting information back then. I thought, this guy really knows what he’s talking about. You had a real technical affinity for Google. And you went through this transformation of having idols at the time, like Perry Marshall, and then becoming co-author with him on pretty much the modern day updated version of the bible of AdWords. What’s that publication called?
Mike: The Ultimate Guide to Google AdWords.
James: Right. You can get that on Amazon.
How do you go about books like that? Can you do Audibles on things that technical?
Mike: There’s an awful lot of screenshots in there. Obviously, things like that go out of print almost, you know, the day they get printed. We only had a small portion of Perry’s audience ask for an audio version. We’ve instead decided to go the video route. So, we supplement the book with a bunch of videos. The book buyers get a whole bunch of free videos as well, rather than go the pure audio route, because yeah, it is quite technical.
James: It makes more sense. I was having the same conversation with a client of mine who’s a program instructor. It’s very hard to do an Audible version, how to code. So, you have been pretty busy lately. You run a website called agencysavvy.com. You have been winning awards. Let’s talk about that for a moment, because I think this really sets up why someone should be listening to this podcast about Google ads. What was your award?
Mike: We’ve won a few this year. It’s the first year we’ve ever – the reason I’m really proud of this is it’s the first year we’ve ever entered for any. I’ve never ever thought much of awards in the past, because they tend to be for creativity and not for results. So, we looked through all of the ones that were available and went after a couple of performance ones.
So, we’ve won a couple with Google this year. And then last week, we were a finalist. Didn’t quite take out the top spot, but we were a finalist for the NORA awards, which is up in Sydney. So that’s the national online retailers. And we’ve had, more importantly, a lot of client success this year. So, one of our clients, Showpo, up in Sydney, won the national retailer of the year, which we’re very proud of. And they had some nice things to say, obviously. They do the vast majority of the work there, Jane and Mark and Alex and the team, but yeah, it’s been great. The team have put in a huge effort, they’re all case study based, so it’s all based on results and performance, not, you know, we did this really funny ad that played for 15 seconds or something like that.
A lot of awards are for like, somebody had a funny ad play for 15 seconds or something like that, and I never wanted to go for an award that was purely a creative award. So getting awards based on performance and based on increasing client’s revenue and return on ad spend, that’s the sort of award that we’re happy to win.
James: It’s funny, I’ve known you for such a long time but I swear you speak more like a robot now than when I first met you.
Mike: I’m not sure if that’s a compliment or not. OK.
James: No, it’s just an observation. I think it’s distinctly different and very precise. When it comes to awards, just as a curiosity point, do you have someone on the team look up awards that you’re interested in competing for, or is there some award expert you deal with who tells you which awards are good to be within your field?
Mike: No, we’re not that fancy. No, we do all of that in-house. We got the amazing Alyssa Anson, who have been doing all this stuff for us for a while. And again, it’s another thing we focused on for the first time this year is our own marketing. You know, it’s always been a case of the cobbler’s shoes, and we’ve never really looked after our own stuff. And that’s been a big focus for us this year. And yeah, results are there already. That’s a really nice thing to take home some of that stuff, first year of trying.
James: Congratulations, and I’m sure there’ll be plenty more coming.
Mike: Thank you.
James: Of course, I made a comment on our social media feed recently that I’ve always known that you were award winners in my mind. I think for the first seven years or so, when I had retail clients through my little agency, you were running the AdWords for me in the background, which was wonderful. And you always did a great job, until I ended up stopping that product line in my business, because it no longer fitted our core. I just want to say thanks for all those years of fantastic reliable service.
Machines beating humans
James: You’ve been traveling a lot and mingling and learning, which is great. You and I bumped into each other in Queensland. We both spoke at an event, and I watched your presentation, which was about AI, which I’m pretty sure stands for artificial intelligence. And I liked that presentation. I was wondering if you’d be able to share some of the latest things going on and I will, like, one little spoiler. The thing that really caught my attention was that you said for about for the first time, the artificial intelligence is able to outperform humans in some of your agency roles.
“For the first time, artificial intelligence can outperform humans in some agency roles.”
Mike: Yeah. So, I guess, well, let’s start with a bit of story, because just to illustrate the power of it. I think most people, particularly in your audience, it’s a very smart group of people. So, they’ve probably heard of AlphaGo and Google’s machine beating humans at this ancient Chinese strategy game of Go. But the very, very short version of that is, as recently as 2014, most computer geeks thought that was an impossible task. That was something that would take decades to do, and maybe we’d never be able to do it, because it’s such a complicated game, and it’s a game that requires intuition. And surely, we can’t teach computers to have intuition, right?
Well, Google bought this company DeepMind in the UK, and within two years, they took on one of the top guys in the world and beat him, beat him four games to one. Then the next year, which was 2017, they took on the best guy in the world, the world champ, and beat him three nil.
Then they built a new machine, didn’t feed it any examples of humans playing the game, which is how they trained the first machine. They just said, here are the rules of Go. And the rules of Go are super, super simple, you can explain them all in nine sentences. But it’s a bloody hard game to master, because, again, it’s intuition and experience. And there are so many possible board positions, I mean, more possible board positions than there are atoms in the universe. So “a lot” is the technical term.
Then they’d built this machine, and they said, we’re not going to train it. We’re not going to teach it how humans have played this game. We’re just going to say, here’s the rules, you figure it out, right? OK, now we’ve got a computer that thinks it knows the rules – let’s turn it on. And within three hours, it’s as good as the average human. In less than a day, it’s as good as a professional human. And in the space of weeks, essentially, it beat that machine that beat the world champ. It beat that previous best machine 100 games to nil. In just 40 days.
That technology is the technology that’s powering an awful lot of Google’s AI behind the scenes. So, it’s unbelievably powerful. But it’s good at very, very narrow things. So, we need to sort of break down what the computer should be doing, what we might want to test, what we should probably keep on the human side for a little bit longer, because computers probably can’t do all of those things. It’s very, very good at narrow tasks. It’s not replacing a digital agency anytime soon. It’s not replacing a marketer anytime soon, but it will replace certain tasks that marketers do.
So, the framework I then created around that is very simple, just to give your listener something to picture in your mind, because what is advertising but showing the right ad or the right message to the right person at the right time, and doing so profitably? So, let’s build a little pyramid, and the bottom layer of that, the foundational layer of the pyramid, has been bidding, doing so profitably. The middle layer is then targeting who’s seeing the ad and when. And then that top little tip of the pyramid is then the messaging, the ad and the landing page and the offer and so on.
Starting at the bottom: machines bidding
So, if we start at the bottom, because I think computers are coming bottom up through that pyramid, bidding is the stuff that they find the easiest to do. That’s just a massive math equation, essentially. And they’ve got perfect data to be able to know how all of us advertisers have performed over the last 18 years, to see every keyword and every bid that’s ever been used. And they have near perfect data on the other side of the fence, which is all of us, all of us humans that are going to Google and searching and clicking on things.
We all like to think we’re special snowflakes, but really, in large groups, we all tend to behave pretty similarly. And Google have an awful lot of data on all of us, because it has things like analytics and YouTube and Chrome and Android and all of these data-collecting platforms that they give us for free to learn more and more about who we are and how we use stuff. So, they’ve got a lot of data. So, figuring out what the perfect bid should be for that particular keyword at that particular point in time, not super hard for the machine.
But Google have been pushing us to use their Smart Bidding for probably the last 18 months or so, and we’ve been pushing back pretty hard. We’re not prepared to adopt everything that Google tell us to adopt just because there’s this new shiny object there. We’re very lucky that we’ve got this big range of clients. Some are more conservative, and some are very aggressive. And so, we’ve got some that want us to test everything as soon as it comes out and test to disprove, if you like. And then we’ve got others at the other end of the scale that are like meh, just roll stuff out to my account when you’re sure it’s going to work, when you sure it’s going to be an improvement.
So, we’ve been testing pretty hard for the last, I’d say, at least three years, maybe three and a half, four years, testing Google’s Smart Bidding against our way of doing stuff. And this time last year, we were winning the vast majority of cases. You know, most stats made up on the spot, but let’s just say 80 percent of the time we are winning that.
“Never believe a statistic you haven’t manipulated yourself.”
James: Never believe a statistic you haven’t manipulated yourself.
Mike: That’s good. I love that.
Yeah, probably beginning of this year, 2018, it was, you know, maybe sometimes the computer would win, and we’d start moving more towards 50-50. And now certainly, with certain types of campaign, the machine is winning more often than not. There’s still plenty of outliers. There’s still plenty of times that you point at it and go aw, but we wouldn’t have done that, or, you know, Google will charge you, we’ve got one case, $367 for a click in a market that should be around 15, 20 bucks. But if I look at that through an engineer’s eyes, I go, well, that’s just an outlier.
Yeah, machine learning is fuzzy. It’s not going to get things right every single time. It’s not using a whole bunch of if-then rules. You know, that’s how we’ve been programming computers for the last 50 years essentially, is a whole bunch of if-then rules. If this, then that. But if this, then do that. Otherwise, do this. Google figured out that they just can’t keep building more and more rules in a really, really complicated, mobile-first, always changing world.
James: Sounds like that Chinese game would have been good experience.
Games and presentations
Mike: They knew that by teaching the machine, or rather having the machine learn how to play the Chinese game of Go, they knew that they’d be able to take the learnings from that and adapt them to all of these different use cases. That’s why they leant so heavily towards the game-playing side. OK, yeah, computers playing games is something that goes all the way back to the 1950s. Bertie the Brain was the world’s first computer game, debuted in 1950 at the Canadian National exhibition. And they’ve been improving things since then. Kasparov got his ass kicked in 97 by Deep Blue. Go was considered the impossible one. But now that they’ve cracked that, some of the stuff that DeepMind is doing is absolutely crazy.
James: Just as a side note there, back in Episode 583, I was talking to James Taylor about artificial intelligence. And he was using Watson to help him compose presentations. Where does Watson figure into the equation, if at all, when we’re talking about artificial intelligence? Would it be something that Google would use as well? Or is it considered a competitor?
Mike: It’s probably considered a competitor. Interestingly enough, IBM refer to AI as augmented intelligence. They’re sort of trying to coin their own phrase there. Another speaker from SuperFast, Nat Husson, has done a lot of work with IBM, probably better qualified to talk about them than I am. But yeah, I don’t believe that Google are using IBM API’s for their stuff. I think they’ve built their own. I think they’re far and away the leaders in AI on the planet.
But AI encompasses a ton of different things, right? There’s robotics in there. There’s language recognition and language generation, there’s image recognition. What we’re really talking about is a little corner of AI, which is machine learning. And even then, inside of that, there’s this little corner of machine learning called deep learning. That’s where a lot of the advances have been over the past five years or so. And it’s deep learning, DeepMind. That’s where the name comes from, which is not to wander off too much down that rabbit hole, but essentially using some of what we think we know about how the brain works, and loosely using some of that structure, which a guy called Turing came up within the 1950s. The theory was there in the 50s.
James: The Code Cracker.
“We create more data every two years than has ever existed in human history before.”
Mike: Yeah, but he didn’t have the computing power to actually create it, which is mental, if you think about it. Like, the guy mapped it all out on paper, worked out how it could work, but didn’t have the computing power. But over the last few years, we’ve been creating more data. I mean, what’s the stat? Like, we create more data every two years than has ever existed in human history before. The stat I heard the other day was by 2020 they reckon we’ll create more data in two hours than existed up until 2013. So, all of human history up until 2013, we’ll create that volume of data in two hours every two hours by 2020, which is reasonably mental. And so, machines like data. So, lots and lots of data plus super, super, super quick machines means that now Turing’s division has been able to be made real. And of course, the researchers just keep getting better and better and better at tweaking their algorithms and making these computers learn more and more things.
James: It’s already overwhelming for the average human. I’m just thinking what this means. You’ve been catching those Black Mirror episodes?
Mike: I love Black Mirror. I’ve actually, I’ve written one. I’m trying to find, if anyone listening knows the guy in the UK that is the creator of Black Mirror, please make an intro, because I’ve got, I think it’s a great episode, but I’m happy for him to turn around and tell me in five seconds it’s an absolute idiotic idea.
James: Well, that one about Nosedive, like, it’s already happening in China. It’s amazing.
Mike: The ratings one?
James: I saw a reporter on social media record a train announcement, and it said you’ve got to obey the train commands or your credit score will be reduced.
James: Like, what?
So, what do you do after bidding? I’m going to rephrase the pyramid into a triangle with the wide part at the bottom. I don’t want to get flagged for the wrong key phrases here, and this deep learning thing. So, what comes after bidding?
Mike: OK, so bidding, just to finish off bidding, there’s a bunch of different models, but I would recommend you test whatever you’re doing now. And bidding, I’ll just assume you’re using it, but you test what you’re doing against one of Google’s models. If you’re a lead gen site, then use Google’s Target CPA or cost per action, cost per acquisition, test what you’re doing against Target CPA. And if you’re a shopping site, definitely test Target ROAS, the bidding strategy there, we’ve moved almost all of our shopping stuff to Google’s smart bidding, testing Target ROAS. So that’s bidding, out of the way. Or it’s getting to the point where it’s almost impossible to beat the machine.
“It’s getting to the point where it’s almost impossible to beat the machine.”
Next layer: targeting
Middle layer is then targeting. So, Google back in 2000 and I want to say, 15, yeah, came up with this way of targeting ads in Apps called Universal App campaigns. And it was rubbish. But two years later, they made that the only way that you could run ads. So, there’s a little bit of form there, just to keep in the back of our minds. They roll something out, they test it, they improve it, it’s still not very good. But one day, it becomes the only way of doing stuff.
2017, they rolled out what they called Smart Display campaigns, which is a way of saying, basically, leave all of the targeting to us. You’ve got this display ad, this banner ad that you want to run, leave it to us, let the machine decide who to show that ad to and when. And it’s pretty crap. Really doesn’t work that well in the vast majority of cases earlier this year, but you can still target your display the old way. So that hasn’t been replaced yet. But again, they’ve got form.
Earlier this year, they rolled out what they’re calling Smart Shopping campaigns. And again, we’ve tested this extensively. Some of our tests showed some really promising early results. Most now don’t, and most, almost, I think, all but one, shopping clients have been rolled back to the old style where we had a lot more control, we’re in control of the targeting, it’s a much more granular approach. The machine just isn’t quite ready yet. I’d liken it to a toddler, right? It’ll do amazing things every now and again, but it really needs a lot of supervision at this point. It will do some pretty dangerous things, too, and unless you keep a close eye on it, it’ll run off into the middle of the road.
AI gone rogue
James: Is it like that one that someone created where it started becoming racist and abusive?
Mike: Yeah, well, AI is really at the mercy of the data that you feed in. And most data that we as humans have is biased in some way, skewed in some way. Amazon had one the other day that they fed in basically all of the resumes of people that they’d, not just people they hired, but people that had applied to them, but that was their data set, and I believe that there’s something like 95 percent of their engineers are male. And so, this machine basically then starts to assume that to be a good engineer, you have to be male, which clearly isn’t the case. But you feed it data that’s 95 percent one thing and 5 percent the other, it’s going to start making these wrong assumptions.
So, there’s a huge amount of human oversight needed at this point to say, is the machine running off on this idiotic tangent that we as humans just immediately know, ‘well, yeah, that doesn’t make any sense’? And there’s lots of cases of those. There’s lots of cases of AI doing “stupid things”. So, I think what’s really, really important is you learn the fundamentals still. You understand how to do whatever the task is – in this case, Google ads – that you want the machine to do, so that you know when it’s running off the rails and you know how to correct it.
James: You could be talking about my robot vacuum cleaner, here. It does a pretty good job most of the time, but yeah, it does some weird stuff. It’s actually quite vicious. Sometimes it’ll try and tip over a surfboard or carry away something on its back. So, I still have to learn how to vacuum, just in case. But it’s good with sand. I’ll give it credit where it’s due. His name’s Bruce, by the way.
Mike: Bruce the vacuum cleaner. Of course. It’s nice.
The human element
So, it’s not terrible. There are some promising signs there, but I just don’t think targeting is completely ready for prime time. And we’re certainly not ready to hand over control, you know? Google will sell it to us as an agency, they’ll say, if you use this more, you’d be able to reduce headcount by 20 percent. Like, that’s not my motivation. Yes, I’d love to see a smaller wage bill, but not at the expense of creating jobs and doing the right thing by clients. We’ve got an epic team. I’m not interested in, saying, “Oh, great, we can get rid of half a dozen people because of blah blah blah…”
James: It reminds me of a supermarket where we get to check ourselves out now.
Mike: Yeah. And I don’t think that makes for a better customer experience.
James: No. Like, when I go to the Philippines, you don’t even have to get out of the car to get fuel. And there’s like, two people bagging groceries, so quite a different experience. Shocking, like actual service.
Mike: I remember running an event in Kota Kinabalu in Malaysia.
James: In where? Bless you!
Mike: Borneo, basically. And we had about, yeah, their team would come in and set up and rejig the main room overnight, and there’d be 30, 40 people in the room. They’re hanging palm fronds from the roof. And you’d walk in the next morning, you’d look in a completely different hotel. Amazing.
James: In the toy store in the Philippines, I counted 50 people in the toy store. There was literally three to four people per aisle.
James: Same when I went to a hardware store to buy a steaming wand, because I had to go to a wedding. There was four people demonstrating a steaming wand versus an iron. They’re like, standing around, they take the apron off, they demonstrate it on the, they call it “kuya,” “ate,” on their apron, and they’re making sure it’s good and they unpack it all and test it all and put it all back and they show you how it all works. I mean, the demonstration levels involved in buying a steaming wand.
I haven’t seen that level of service in a long, long time. You go into Kmart or Target or something like that in Australia, you’d be lucky to find one person in the entire store.
Mike: And they probably don’t know what are the things.
James: They have no idea what’s going on, if you do find them.
Mike: Yeah, yeah.
Mike: And a lot of the time, yeah, Google is essentially engineering and sales. If you’re an engineer, you come up with this idea. You run your tests, you model it, but ultimately, you need it running in the real world, collecting new data, to see if this is going to work. And yeah, some of the stuff they come up with just does not work. But then they’ve got to go to sales and sales have basically got to come to us agencies and say, “Right, we need you to do this” And maybe there’s some incentive there, maybe they’re bonused on that, I don’t know (I’ve been told categorically they’re not) but, yeah, they do seem to focus on adoption a lot these days, rather than just…
James: Well, just don’t mention “wave” or “buzz” or “plus”.
Mike: Right. It would just be revenue, right? They would be bonused on growth or revenue. Now it’s a bunch of different things. Adoption is a big, big thing. But again, we’re going to test that on some accounts, and we’re very lucky – we get to see the results of that across hundreds of accounts, thousands of different campaigns, where it works, where it doesn’t, which situations it might work in. And if you’re in-house or you just work on a couple of accounts, you don’t get the luxury of that. So, we’re very, very lucky in that regard. But some things are just not ready to be rolled out, and you don’t want to be first with this stuff. You know, pioneers get arrows in their backs. An early second or third sometimes makes a lot of sense, and sometimes it makes sense to hang way back and let other people go while they improve it and improve it and improve it.
“Pioneers get arrows in their backs.”
James: You know, to listen to Mike Rhodes on a podcast is a very smart thing to do.
So, what’s after target?
At the top: messaging
Mike: Then, messaging. Let me just check in one that is working well, just on targeting if we’ve quickly got time. It’s a thing called custom intent. So, there’s a thing called In-Market. Let’s go there, first. The thing called In-Market, there are 500 or so fixed categories that basically say, if somebody is in the market for, let’s say, a brand-new Audi, or for SEO services, or for a payroll system, there’s all these different categories. And you can find a list of them on Google’s website. But they’re fixed. And there’s only these 500. It’s wonderful. Google are looking at your search history, your browsing history, you know, what you’ve been looking at on YouTube, to determine what you’re in the market for.
So, this abnormal behavior, because normally, maybe you look at baseball sites and golf sites all the time, and then all of a sudden, you’re looking at washing machines and you’re searching for a new washing machine. Google spots that, says, aha – Mike’s in the market for a new washing machine. You can target your ads based on that, which is wonderful, if you sell washing machines. But if you sell something that doesn’t fit neatly into those 500 categories, you’ve got nothing else to do, until recently.
And then they brought up this thing called custom content, which really should be called custom in-market because that’s what it is, but it’s called custom intent. So instead of having to pick one of these 500, you get to build your own. You get to build your own theme and say, hey, people that are into this, people that are in the market for this, people that have recently been searching for and browsing information about washing machines, let’s just say, I want to show ads to those people, please Google. So, it is not – I have to stress this – it is not the same as putting your remarketing code on your competitor’s site. But it’s about the next best bloody thing that you can get, right? Because you can say to Google, hey, people that have just been browsing sites like this and searching for terms like this,
put my ads in front of them, please. So that is one case where that information is not available to us as advertisers. Even if Google wanted to give us that amount of data, they couldn’t. It’s just too much data there. So that’s where you do have to sort of lean in, trust the machine, and that’s well worth testing.
And I think that’s the future of Google Display. I think they’ll do away with keyword targeting over time, and customer intent will become the only way, or the main way to market there.
James: That’s super relevant, isn’t it?
Mike: Oh it’s… and Google knows, right? I mean, if you can now show an ad to the person that is looking for the thing that you want to buy, rather than the old spray-and-pray approach.
James: It’s just like I was speaking to someone about the difference between boosting ads vs only advertising to people who have watched their videos. I was talking about Facebook, but it’s quite a different intent there. And that’s the difference between a complaint or an order.
What technology is learning about us
James: But yeah, I mean, I’ve paid attention to how much these technologies are learning about us. When your phone tells you how long you’re going to get till you arrive at your destination, which you never told it, but it’s figured out you do the same trip over and over again, that’s like, OK, it’s really a lot going on here.
Mike: Or, I was on a podcast the other day with Michael Stelzner. We were talking about Social Media Marketing World earlier. His phone popped up a notification for him saying, looks like you got a meeting with Mike Rhodes coming up. Do you want to call him? Now, he had not even put my number into his phone. But I’d been swapping information with his personal assistant. And in one of those emails, my phone number was mentioned. The AI’d figured all that out, put two and two together, come up with four and a half and said, you might need this in a second.
James: I was going to ask you about that. I noticed in my Gmail, it’s now giving me suggested responses that seem incredibly accurate.
Mike: Why, so it’s started doing suggested responses. You get those three responses down the bottom, and I think something like 20 percent of people use those on mobile. But now it’s starting to finish your sentences for you.
James: It’s definitely finishing my sentences with words that I would actually use.
Mike: And the more you use it, the better it’s going to get.
James: And the responses are not the same responses.
Mike: Oh, no.
James: I saw it in LinkedIn, as well. But Gmail is doing some pretty crazy stuff right now for me.
Mike: So that’s a perfect segue into messaging, right?
James: Oh, I love it. Great.
Can AI write your ads?
Mike: Well, computers can learn to write. And I use Gboard. I’ve been an iPhone user for 11 years, I have resisted going to the Pixel 3. I’ll probably resist a little bit longer. But I pretty much Google-fy my phone, I use Google Maps, I use Gboard, I use Google Assistant on there, Google Calendar. It learns words that I use. It’s just, it’s crazy.
Inside of Google Ads, it will try and write ads for you, but it will use information that you’ve given it. So, it’ll look through your landing pages, it’ll look through the ads that you’ve written in the past. And it’ll start writing ads for you. And you know what? They’re actually not too terrible these days. A year ago, they were awful, and I would say, run away from them. But these days, have a look at them. We tend to turn them off in client accounts. We certainly turn off having them automatically be used.
So, Google will pop these up into your account, they’ll sit there for 14 days, basically paused for you to review. And then if you don’t do anything, they’ll start running. So, there’s an option to turn that off. But have a look at them – they’re not terrible.
But you know, we’re not at the point yet of the AI writing a great landing page, you know, persuasive copywriting or crafting an offer. We’re definitely not at that stage yet. And that’s an area where a lot of agencies could add a ton of value, is offer creation, creative writing, creativity in general. You know, coming up with, it could be that banner ads. You know, the amazing Tom Breeze, good friend of both of ours, comes up with the most wonderfully scripted and edited YouTube ads. I mean, the best I’ve ever seen in the world. That isn’t going to get replaced anytime soon by a machine. But writing a little 30-character headline, well, you know, the machine isn’t terrible at that.
And where the machine comes into its own with the ad stuff, instead of now – I know you’ve seen the extra big ads, so we now have the option of using up to three headlines and two description lines, virtually doubled the size of an ad a year ago, pushing all of that organic stuff further and further down the page, (thanks, Google, the SEO guys are thanking you, I’m sure. So, we’ve got that option of having three headlines. Well now, with Google, you can enter 15 headlines and it will choose which three to show, and enter four description lines and choose which two to show. And over time, the machine learns which combinations work for which type of people. You don’t get a lot of control over that, you really do have to sort of lean in and trust the machine a bit there, but early results for that have been pretty good, and they’re only going to get better, right?
The exponential curve
This is the thing that we tend to forget with exponentials. All of this is on an exponential curve, which looks like a flat line for a really, really long time, until it’s not, until it goes racing up past you and heads for the moon.
“Exponential curves are not how human beings think.”
Exponential curves are not how human beings think. We remember the past as being just like today, but a little bit worse. And we think five years from now will be pretty much like today, but a little bit better. We forget just how different the world was five or certainly 10 years ago. Five or 10 years from now, it’ll be vastly, vastly different, and you want to stay in front of that. You want to stay agile and keep your wits about you and keep up with this stuff, because it is going to change dramatically. And I think what the machine will be able to do this time next year. If we’re having a chat this time next year, what the machine will be able to do will knock your socks off. It’ll be amazing, and we’ll be thinking, Wow, has it only been a year since? It was not very good at this thing over here and couldn’t really do that at all. It’ll change fast.
James: Yeah. I mean, even in real world stuff. I was just noticing, I have this fantastic coffee machine, but it’s been a bit unreliable, and I replaced it with a compact machine that’s automatic, that’s actually pretty good and very fast and super awesome. It’s new. And then I noticed I’ve replaced my old speakers with a Google Home, so an Apple pod, one of those. I got one of each, actually. I’m sure the CIA or whoever’s listening in perfectly at my house. And, you know, what we can do on our phone now, compared to a few years ago is pretty amazing. Filming cameras, threw out the old DSLR and mics and light and all this stuff. The phone is just incredible with the resolution. And this GoPro that I’ve got with stabilization, it’s what they would have used to make movies 10 years ago.
Mike: Or a Google Translate. I don’t know if you use this in the Philippines. Tom and I were actually both speaking at this event in Moscow earlier this year. So, I’m walking around Moscow on my day off, getting a bit lost. Google Translate, you can hold the phone up to assign it a foreign language, in this case Russian, and it not just recognizes the font and then recognizes that this is Russian and translates it into English. It – in real time, mind you, removes, on the picture that you’re looking at on the phone, it removes the Russian words, it puts the English word onto the sign in the original font that was on that sign, and translates it. And it does all of that in airplane mode. All of that is happening on the phone in real time. It is phenomenal.
James: Oh. I had no idea about that.
Mike: It’s crazy good. And the Google Pixel buds…
James: Now I’m going to know what I’m eating. Finally, the answer. Does it work on menus?
Mike: It does. You can point it at the menu and see in real time all of that translated into English.
James: I’m going to be trying that straight after this podcast. Love it!
Mike: Yeah, it’s crazy good.
Mike: It’s a bit scary. I get that. People are freaked out by it. And is this going to replace us? I don’t believe it replaces jobs. I mean, some jobs maybe. But the vast majority of jobs? Probably not. But it replaces a bunch of tasks. Yes, I don’t think any agency should be sitting there manually tweaking bids, four days out of five. Like, if you’re still doing that at the end of 2018, then either change really, really quickly or you won’t have an agency by the end of 2019.
From now till April
James: You know, I’d be interested to know, what do you think will change between now when this podcast comes out round about the beginning of November to the beginning of April, when you’re speaking at SuperFastBusiness Live in Sydney? What do you think could possibly happen in that time period?
Mike: I got to be careful here, because I know about some of the things that are coming out, but I’m under NDA and can’t talk about that.
James: Aw, you foiled my plan.
Mike: So yeah, that’s really difficult for me to answer.
James: OK, let me phrase this differently. Do you think there’ll be significant leaps forward in some areas between now and then?
James: Would you come and share them with us at the time?
James: OK. Well, that’s good.
Mike: Much easier to answer. Thank you.
James: I’m excited! So, in this general topic, we’ve talked about AI, we’ve talked about bidding, targets, messaging, which things robots are good at, which things they’re still learning. We’ve talked about weird and wonderful things to the side as well. Tours to Russia and so forth, and robotic vacuum cleaners.
What should I be asking you before we wrap that’s critical to know on this topic?
Suggested focus for 2019
Mike: I’d say, what should I as an agency owner – which a lot of your audience is – what should I be focusing on as we head into 2019? And my answer to that would be, what we have focused on as an agency for the past three and a half years, which is three, maybe four things. Those three things are strategy, creativity, and relationships. And then that fourth thing is automation. Automate as much of this as you can, where it makes sense to do so, where client results are not getting worse. I think most people, as much as we’ve been talking about AI, we’ve been talking about the AI that we get to use. It’s built into the machine.
(Clunking sound in the background)
There goes a surfboard.
James: Dead set, that is Bruce.
Mike: That’s hilarious. “He’s going to talk about me. I’m here.”
James: He’s come alive, on the hour, and he’s wreaking havoc in the house.
Mike: That’s hilarious.
James: I think that’s just the boogie board, by the way. It’s nothing too severe. I can go and stop him, one sec.
Mike: You heard it here first, folks, live on air, a robot takes over and tries to go surfing on James’s surfboard – unsuccessfully, because it’s now being turned off and thrown over the balcony. He’s coming back, eventually.
James: And I’m back.
Mike: There we go. That’s quite funny.
James: The wonders of technology just managed to topple one of the surfing craft. Good on you, Bruce.
Mike: So, I don’t think the vast majority of businesses need to worry too much about doing AI. Like, most businesses, yeah. The vast majority, 99 percent of businesses do not need to go out and learn Tensorflow and hire five data scientists and build their own neural networks. You can if you want, but it’s probably a waste of time. There’s going to be AI off the shelf. And that’s something that will change over the next six months. There will be more off-the-shelf AI, it’ll become like Lego blocks, but we’ll have these tools available that we can take off the shelf, put into our business, bolt them together in different ways to do the things that we want to use them for, to automate those mundane manual rote tasks, so that we can focus more all of us on doing the more creative, more strategic work, building better relationships, stronger relationships with our clients.
So, focus on automation, more so than like going in to build AI, but use tools that have AI in them, like Google Ads or Facebook or something like that. Carl Taylor, another good friend of ours, is leading the charge there in terms of automation and getting businesses to automate as much as possible, because we shouldn’t be doing that stuff as humans.
Should you be doing this yourself?
James: Speaking of shouldn’t be doing stuff, if you’re a business owner, is it harder or easier than it used to be to be trying to do this Google Ads yourself versus hiring an agency? I know that’s a super stacked, biased question.
I gave up trying to figure it out a long time ago, about eight years ago, and just let specialists handle it, and I do the same with Facebook. I don’t spend a minute thinking about it, because I feel like it’s so specialized that the difference between someone who knows what they’re doing and someone trying to learn what they’re doing seems like a vast canyon.
Mike: I think the answer to that would depend on your perspective, certainly. If you were in Google sales, you would say it’s much, much easier now. They’re trying very, very hard to simplify, and they know that they need many, many more businesses using the platform, and so they’re constantly trying to dumb down the platform. They’ve brought out this thing in July called Smart Campaigns. Not to be confused with Smart Shopping and Smart Display. Smart Campaigns, which is basically just, give us your credit card, plug in your website, we’ll do the rest, essentially.
And for some businesses, look, maybe that’s OK. For those of us on this side, that know what great results look like and that aren’t happy with OK results, then it is vastly more complicated. There are many, many bells and whistles under there, and look, I wouldn’t want to be learning it from scratch today. I’m very happy. I learned it back in 2004 when it was a much simpler beast. And we’ve just had to evolve our methodologies, our processes as an agency. It’s been much, much easier to do it that way. If you want great results, then I think you want a specialist. I think that’s probably the case with most things though, right? It’s easy to get an OK landing page, but if you want a really good one…
James: I think so. I think you usually start out trying to do stuff yourself, and often it’s because you don’t have the funds, and usually that’s because you haven’t got an offer that converts. But the difference that I’ve seen with my higher-level students, the ones doing millions of dollars per year, they’re not hands on with most of the things like this, unless it’s their thing. And you know, people like you and Tom, they’re my experts who I’m talking to frequently, who are experts at this, but usually they’re good at something else and they get someone like you to help them. I think that’s smart.
Mike: I think it’s a Dan Sullivan phrase that I refer to a lot, “Delegate everything but genius.” So, stick to what you’re really, really good at, and have the recipe outsourced or delegated. That might be to other people in your business, or it’s probably to other businesses that can help you, whether that’s a digital agency or some other use. I mean, I’m just picturing out the difference in, say, a landing page between an average copywriter and somebody like Kevin Rogers. I mean, it’s night and day, right?
So, it depends on your perspective though of what you’re willing to accept and what you can compare that to. If you’ve never seen a landing page before, then yeah, something that the new student knocks out in a couple of days might be, if you compare that to what you could do, maybe you think that’s awesome. But if you compare that to a genius, it’s going to look pretty horrible.
James: I saw one the other day that made my eyes hurt. And it’s not making too many sales and I could guess why. It’s that catch-22. It’s not going to make sales so he can’t reinvest in proper design and words. So, he’s not going to make the sales to reinvest in the proper words and design. So that’s a bit of a tough phase to go through. Once you get past that though, such a great environment to be in.
Mike, I want to thank you ever so much for coming and sharing generously again, and I’m looking forward to finding out about what’s new in 2019 when we catch up face-to-face in Sydney at SuperFastBusiness Live.
Mike: I very much look forward to being there. Thanks man.
Keep up with the latest business developments inside SuperFastBusiness membership
Liked the show? Enjoy all the episodes by subscribing on iTunes