3:2 The real impact of AI on event marketing

Lee:

Welcome to the Event Martech Podcast. This is your host, Lee. And on today's show, we have the one, the only all the way from Canada. It's Alex from Clickmakers. Buddy, how are you doing today?

Alex:

I'm doing amazing. It's a great morning here. We're being eaten alive by mosquitoes this week. But other than that, the summer's going great.

Lee:

That sounds horrific. I went to Boston once and got one mosquito landing on my arm, and I was there in the Fetal position. We just don't get that here in the UK. They scare me. Anyway, for folks who don't know who you are, buddy, could you just let us know who you are, where you're from, a bit about your company, and also a little bit about yourself, maybe something that people don't necessarily know about you.

Alex:

I am the founder of Clickmakers.io, which is a digital media buying agency. We're on traffic across many, many platforms. Our jam is Google. We're a premium partner, which puts us in top three % of US agencies, whichever the Yoast or Google is using for that, I'm not sure, but that's what they say. And that means we get access to their betas. We get access to using the office for events in Redwood City. So we get some perks from that. We can peek around the corner and understand where things are going as far as all things, AI, data, all those Google stuff that they're working on. A couple of years ago, we launched another company that is all about data, data integrations called Measurebit, so measurebit.com. This is where we wrestle Frankenstein monsters and put together Humpty-Dumpties that most websites are those days, where data needs to go from point A to point B, and very often doesn't. We solve those issues. I live about 90 minutes north of right now in a rural area on two and a half acres of forest. We have some turtles in the pond. They come out and they're awesome.

Alex:

But don't play with them because they can snap your finger.

Lee:

I can imagine. So, Alex, you are all about data. It's clear. When I was reading through your website, could you share a moment where data has significantly influenced the marketing decision that you've made?

Alex:

Pretty much every day. I will tell you a story how terrible we are as humans. Maybe it's just my personal talent, I don't know. But how terrible we had predicting the outcome of a particular test. We think we know what's going to happen as experts. I've been doing this for a very long time, so I think I know what's going to happen. This version of the page is going to work, or this version of the ad is going to be better. I was at a software conference a couple of years ago. Well, actually, I was quite difficult of it, so that was more than a couple of years ago. There was a booth where a company that does conversion rate optimization on your site, they would give you a tablet with a bunch of slides on it, and you had to guess that they would give you two versions of a page. You would have to guess which version performed better. You would go through this quiz. There's 10 things on it, 10 questions, and I scored a perfect zero out of 10. I won the prize that day. I was wrong every time. Basically, if you would bet on the opposite of what I thought was a winner, you would have done really, really well.

Alex:

But that just tells you how bad we as humans are at predicting the outcome of any test. And that's why we need to use data and data driven decisions to really understand what's going on, to get to the truth.

Lee:

So you did highlight there the human aspect. I'm imagining too, there is the human aspect looking at the data and still not necessarily being able to ascertain what's going on. How have you seen AI change the marketing landscape over the last few couple of years or so?

Alex:

We've been working with AI or some version of machine learning, which I think is the proper name for it. Ever since Google came out with Smart Bidding, or all these digital media networks started to use in auction, real time bidding is the proper name for it. We delegated the most important task that we used to do just a bit over a decade ago as media bias. In some cases, we still do it. It depends on the network. But most places like Meta, Facebook, Google, Microsoft, the artist formerly known as Bing ads, you have this task of figuring out which auctions to enter, because every impression online is a real time live auction, and how much to bid inside this auction. So you get to decide the value of the bid. And that has been the area where the machines showed up and took over this job very quickly because they were so much better at it for two reasons. A, this requires real time decision making. When someone opens your web page and there's an ad on that page, or when someone loads your ad, maybe it's a Google search results page or maybe a YouTube feed of videos, and there's some ads in there.

Alex:

Every time someone's loading a page on the phone, on their computer, all these ads, they're running real time auctions, like in milliseconds behind the scenes, figuring out which advertisers will be admitting to the auction, and then which advertiser is going to win this auction based on literally thousands of signals. We as human media bias, we are not there. I cannot be there inside millions of auctions taking place every second. Also, as a human media buyer, I don't have access to all the data that the machine has access to for a number of reasons. But the machine knows whether I am sitting at home in front of my computer with my wallet out, I can go out to swipe my credit card, or if I'm paying for my groceries, standing in doing a queue and just killing some time on my phone. It's the same me. Demographically, Psychographically, it's the exact same person. But contextually, it's a very different value of an auction, very different value of that impression that an advertiser would want to buy. The machine is the best way to make this decision, is using the machine, not using a human media buy, but using the is much better in this approach, in this particular case, and it's much more efficient.

Lee:

You say that's been going now for several years as well, as opposed to AI that we know it with ChatGPT and content generation being fashionable for the last couple of years. That's a form of machine learning that's changed the marketing landscape over the last 10 years.

Alex:

It started slow, but then it gradually took over more and more components, right? Started with bidding, then it spread itself into targeting, like all this audience stuff, a lot of that became the area where machine learning has become more prevalent. Now the actual ad creative is where the machines are playing a huge role. They're trying to deliver on the promise of showing the right ad to the right person at the right time. Well, we're very close to getting there now because essentially what the loading into the machine now, like when we launch a campaign, in most cases, it's not an ad per se, it's just a bunch of spare parts. We just load what they call assets. We give them headlines. Here's a bunch of headlines. Here's a bunch of descriptions. Here's a bunch of images. Here's a bunch of videos. The machine will go in and each auction will compile an ad in place, like using the assets you provided. It will try to create an ad that is the most fitting that space in that situation and that person, like what this person is trying to accomplish. We'll know where this ad has been shown contextually, and to whom it's being shown and what this person is doing right now.

Alex:

So yeah, advertising has always been trying to get to the place when each ad is personalized and is the most relevant to the person looking at this page or that medium, whatever that is, be it the newspaper, TV screen, or podcast audio, or search results, mails, email. All those things are different types of media. But the job of the machine right now is to figure out what ad is going to be the most relevant, the most appropriate here, the most effective. Yeah, again, our creativity happens when we figure out those assets, but once we load them in there, then we just hope that we've trained the machine correctly to make good decisions on our behalf when we are not around. We're not around pretty much always. This stuff literally takes Milliseconds now.

Lee:

Where's the human aspect versus the machine aspect? I'm trying to work out the separation.

Alex:

Sure. I know it's a very important question, and I think what I'm going to say is going to be very different than what Google engineers might think. Our role as humans in this marketing process is expanding. It's not shrinking, which is becoming way more productive. It's all the icky, all the mechanical, all the boring things that are being taken over by the machine for the most part. Then we end up doing these things that are the most creative. They are the most human, if you will. I'll start with things like insight. The data tells you what's going on, but it doesn't necessarily tell you why. Yes, the machine has access to literally thousands of signals inside this auction, and it tries to make a decision the best it can. But at the end of the day, still we have played a role of a decision maker as a human operator of the machine about the direction where we want to go. When we get the data, that gets for us to understand maybe the website was down and the machine doesn't have access to this bit of data. Maybe the weather pattern was different, or maybe there's a shift economic conditions, which is a lot of businesses experiencing right now.

Alex:

Again, the machine won't necessarily understand that. Stuff like this, why this campaign is performing worse today than it did same quarter last year? Well, there could be a number of reasons that the machine won't understand and won't know. We're just saying, Well, this campaign doesn't perform as well as it used to, is not good enough of an answer. We have to drill down and we have to dig deeper. Strategy, the machine is not going to come up with anything really. The machine can just twist the Rubik's cube and come up with a different set of colors, or it can just assemble the jigsaw puzzle in a different way every time. But the pieces to go into the jigsaw puzzle and the goal we're trying to achieve, that's going to come from us. The overall strategy that comes from a human still and for the foreseeable future, I think that task will remain with us. The third thing I would like to highlight is relationships. I think as more and more AI and deep fakes and all those avatars, they will become prevalent. It's scary where we're going a little bit, but also exciting. I think the amount of content that's going to be generated by AI is going to explode.

Alex:

We're already living through it. It's already happening. But as I say, you ain't seen nothing yet. I think there's going to be way more of this. I think it will level the playing field for a lot of people from other countries that couldn't speak English without a thick accent, for example. All of a sudden, they can become as good as TV announcements.

Lee:

Yeah, I've seen that. That's incredible, isn't it?

Alex:

Right. You know, someone who has a face great for radio, all of a sudden, they can use an avatar to sell something to an audience they didn't have access to before without necessarily hiring talent to become that avatar. I think there's going to be a major jump on productivity and a major jump in us accepting that we need to interact with some version of AI, be it voice-generated AI, the chat interface or something like that. But human connections will become even so more valuable with all this stuff. Just when we could photocopy the Mona Lisa, the value of Mona Lisa did not decrease. People still go to the museum and wait for hours for a chance to see the original. The original, the human will rise in value. Us talking like this on a video like human to human, that's going to be more valuable, and that's not going the way that I think that's only going to appreciate.

Lee:

I think the example you gave about why was the campaign different, say, yesterday or last week versus today, is very relevant because today in the UK, as an example, the analysts were shocked that the rate of inflation actually dropped, and therefore they now have to change their prediction with regards to what the interest rate is going to cap out at, etc. All of a sudden, everyone was wrong. All of the certainties that everyone was projecting have not happened. Then suddenly there are new projections coming out, etc. People might be making, as of today and maybe the next few days, new buying decisions based on some slightly more positive information that's come out. Like I said, the AI isn't necessarily going to know that straight away. That's not to say that AI might not know that in the future and they might be able to extrapolate even things like that, but then you're missing, then you still need the human element, the strategy element, or even just the element of knowing when to just pull the plug for a while. Something's hitting the fan globally. Let's just pull the plug just for today, and then let's revisit tomorrow.

Lee:

Just all of those human elements. I think what you're saying here is that AI will become tools just like a pen and paper. These are tools, and this will level the playing field for people, and we'll see an explosion, really, in creativity through all of these tools. But with that aspect of humanity, humanity being so freaking important. I like the idea that you and my and your interview right now will be as valuable as the Mona Lisa, mate. That's beautiful.

Alex:

I wouldn't necessarily go as far, but let's refer it. You never know.

Lee:

Why not? You never know. We might be framed in one day.

Alex:

Hey, I'm the worst predictor of the outcome of an A/B test.

Lee:

There you go. We're looking here, we're making data-driven decisions. We're using machine learning, we're using AI. I'm still struggling to work out the creativity part. Back in the day, say, 10, 20 years ago when I was doing campaigns, so design to print campaigns, for example, for an agency, we would always come up with a whole load of different types of messaging, and we'd come up with our own designs and imagery, etc, and we felt pretty free to do anything. We didn't have the data. We'd be pushing things out into magazines or on billboards, etc, But there was an awful lot of creativity and excitement. If you're throwing in AI and data into the mix and you're filling in a few check boxes for wording that could be used or a few images, etc, Because that's how you were describing it earlier on, where can we still be creative? And how can we balance creativity versus data? Especially if data is telling us one thing, but we really want our brand to shine through, or our brand voice to shine through as an example.

Alex:

There are several things to unpack here. Sure. Number one, Rory Sunderland, who runs the Ogilvy division of European Office of Ogilvy based in London, near you.

Lee:

Yeah.

Alex:

And he says, The opposite of a good idea of a great idea is another great idea. So the AI is not going to be able to figure out the opposite of a great idea. That's also a great idea. That's something that humans will have. It will take a human to show up and be human. I can say, No, I think we got to test this. Then you let the data tell you how good your test was or wasn't. If this new direction has a chance of being successful, defining this direction creatively. Let's try something completely off the wall. That's just something crazy, right? That the AI is not going to suggest that necessarily. The AI we know today, again, that's not to say that the AI is not going to evolve and become something bigger than it is today, but the AI today allows us to build images. It allows us to edit videos much faster. The turnaround time, so the lead time from conception to launch, is being shrunk significantly.

Alex:

Our productivity is going through the roof, which I think is amazing at a time when all countries become a super world. Most major economies are about to hit a concrete wall demographically, and there's going to be just not enough people to do all the jobs that need to be done. The AI showing up right now, like when we need more productivity, we need this huge jump in productivity. When things are shaking economically and the jump in productivity is the only way we can pull ourselves out of a hallway, this economic abyss we're staring into right now, I think it's just a marvelous thing. I think if you allow me to get a little superstitious, this is a godsend. It may be. I think the timing is very interesting. We need a jump of productivity right now, and the AI is enabling that. As far as creativity, first of all, creativity thrives when there are some constraints in place, like when there's direction. So being data-driven enables those constraints for you to be creative within. Having no constraints is not good for productivity. For creativity, for creativity. It's the blank page syndrome, right? What are you going to write about?

Alex:

You're staring at the blank page right now. Give yourself a topic, grab a question from a forum, right? Give yourself some type of constraint. That's what being driven enables us to do, but also it allows us to go in the opposite direction. Here, we tried this. How about we try something very different? AI will try to stick to what's been working in the past because it only has hindsight. It has no foresight. It's foresight. It's only based on the data that it has. Using a mountain climbing metaphor, you've been climbing this side of the mountain, and you are not succeeding. Well, maybe you should go back to the base camp and start climbing the other way. The AI is not going to do that. That's the job of a human to be creative and say, Well, we're not going to reach the peak this way. Let's go back with group and try something completely different from now on. It's the same. Yeah. Again, creativity is only going to go up, in my opinion, because now everyone can be a writer, everyone can be poet, everyone can be an artist, everyone can be a video producer.

Alex:

And you don't even need to be good on camera or have a radio voice to succeed most of the time. Unless there's a requirement to have a human inside that transaction, you can do a lot of things with AI, or will be able to very soon.

Lee:

Now, you guys focus on driving traffic to websites, I believe. What are some common mistakes you've seen with businesses trying to drive traffic to their websites?

Alex:

Well, the number one mistake, number one, in terms of being the most common, is not having proper conversion track in place, not being able to measure the results properly. That's number one. Number two, having conversion track in place, but measuring the wrong thing, optimizing for the wrong goal, or using the last century mentality, last entry math to the 21st century AI-driven marketing. I'll explain what I'm talking about. In the past, let's say you have a transaction value of 2,000 pounds. Let's say you need 10 leads to get one order. What's the value for the lead? Everyone's going to go immediately 200 pounds. The correct answer is today, and has always been, but the math used to work out, but doesn't anymore explain why. The answer is you have one lead worth 2,000 pounds and nine leads worth exactly zero. The only problem is you don't really know which one is which. That's why you apply this law of averages, or you do this basic arithmetic to figure out the average value of the lead. Well, averages don't work very well anymore. The reason for that is AI. Many networks, many campaigns are becoming multichannel. On Google, the major type of a campaign we run these days is called Performance Max.

Alex:

It's new. It showed up a year and a half ago for the first time, and has been evolving. But that's where Google is going. If you launch something on YouTube, let's say you launch an ad on YouTube, and you think you buy an instream, meaning that someone's watching the video and there's going to be your ad before in the middle of somewhere, interrupting that video. Well, that's how things used to be two years ago. But since then, YouTube has become a multichannel platform. Now, when you're launching a YouTube campaign with a conversion objective, it's going to buy some of the instream inventory, some of the in-feed inventory, meaning it's going to be a banner in the YouTube feed. It's going to buy some from Google Discovery and some from the display inventory, meaning it will be banners or videos shown inside banner placements on content websites, on the entire Google Display network. That means that not only does the machine make a decision about which auctions to enter within the same network, it first decides which network to grab this impression from. The value of this impression is dramatically different. Or even if this person who clicks on an ad, and very often it's not going to be a person, it could be a bot, or could be a real person, but from a click farm, there's a lot of spam online, as we all know.

Alex:

What's happening now is you grab this impression from all the machine buys some impressions from the display network for you. These people from the click farm, they fill out your forms and they register for you event. Let's say it's a free registration. If you're taking money, you got to be pretty much immune to this stuff. If you're charging money from the get-go. But if it's a free lead gen type situation, like if it's an opt-in form or free registration, very often when you launch campaigns on YouTube, on Performance Max, on Facebook, with Facebook audience network expansion on any social media network, ad network with the audience network expansion, you start getting those fake leads, not just fake clicks. You start getting fake leads. Very often those are not even bots. Like I said, those could be real humans just putting the fake info.

Lee:

Yeah.

Alex:

The way to mitigate this, the goal is fake click solutions. They're not very effective. You can try, but usually they're not effective. The approach that we use for that is that you break the link between reporting that conversion back to the network. Instead of saying, Hey, Google, thank you. Just made me 200 quid. That's not true. You first look at that lead, you try to qualify this lead. You can immediately spot if those are fake email addresses. You can kick off an API call to a third-party provider and verify that this is in fact a real person. Maybe you can get something like the credit score. Based on the geography, I'm not up to speed on the UK privacy legislation, but they're providers that will expand your data set by adding more columns to it and telling you, Yeah, this is a real person, or this person is a known complainer and refunder. What you want to do is when you get those negative signals, you just don't fire this conversion pixel or conversion event for Google or for Facebook or whichever platform you're running traffic on, because you don't want to reward the AI for doing something wrong.

Alex:

You don't want to give a dog a tree to puppy that made a pile on the carpet in the middle of the living room. Yeah. Because he'll keep doing it. It's much easier to make a pile right here than to ask you to let them outside when there's wind and rain and cold. You want the machine to keep going into high intent, but hard to win expensive auctions to buy quality clicks to bring you quality traffic. You don't want it to keep going to like spam infested, bot infested, and ClickUp infested inventory. That is all fake. That's the major component of what we do on our projects. Most of our media bias, they have engineering degrees of different kinds. They're very technical, they're very data driven. Then we have Measurebit, which enables this stuff that enables us to build this lead qualification or lead scoring mechanics inside those funnels that we built for clients that allow us to run successful media campaigns.

Lee:

That's incredible. I mean, Ithink that's certainly the mistake I've made in the past is just reporting all of that data straight back. I am giving this positive feedback loop when actually about 70% of one campaign I did was actually just spam emails. It was a free to sign up as well, so there was no payment via wall. This was data going straight in. This was training the algorithm that what we were doing was good, and we were ending up actually getting a random audience from Asia. I think it was bots, etc, that were just filling in forms, and we were just getting all of this data with fake emails that we could do nothing with. But we were spending money. We'd left it on autopilot. We hadn't looked at it for a little while, and we were like, Oh, my word, this was a complete waste of time. At the time, I didn't really know what to do about that. Everything you've just said has just blown my mind.

Alex:

Well, let me just give you another metaphor. Sure. This is something that's changed recently, maybe in the last year and a half, two years. It used to be that you would launch a campaign like this and you would get some spam and you would just account for it. You were like, Yeah, we get some good ones and some bad ones. But you just learned, Okay, so we get the bad ones from this targeting from this audience. They just not run this campaign anymore. More. This loop where you decide what to target, launch a campaign, get the data, optimize, launch, get the data, this loop is now instant. There's no break in there for you to step in and make a decision and adjust the direction of the campaign. That happens instantly. The machine learns on the data that you send right away. It's more like a cancer cell in the body. It tries to... It affects all the other cells around it and then tries to stage coup. This is how your campaign becomes... It learns bad habits, and it's very hard to get it to unlearn this stuff later on. That's why you want to, from the get go, you want to make sure that there's some type of mechanism for qualifying the that are coming in.

Alex:

In sending much cleaner signal. You want more signal, less noise going back to the machine, because that's the only way the machine learns whether or not it's doing a good thing, right? It's like training a pet or even a human, like any type of intelligence we know in the world. That's how it's trained through reinforcement, right? In machine learning, it's called reinforcement learning. That's the actual term they use. That's what it is. You reinforce the behavior. If the machine behaves badly, you don't want to reinforce that.

Lee:

So the era of set and forget is not true. It's set, monitor, qualify, review, adjust, update.

Alex:

And keep going. It's more like you have, again, the pet training metaphor, the dog training metaphor, you want this dog to be able to walk near you off the leash and not chase the bicycle. That's what you want. We used to be able to hold tight to that leash. That's the way we used to manage things. He would try to chase the bicycle, but he would just pull it back. We don't have the leash anymore. You have to train the guy so that he's not going to chase the bicycle on his own. You're not going to be able to do anything. You just train him. That's how you're going to impact this campaign. And then let him live in the world and make decisions on his own. Make good decisions, hopefully, based on how you train him.

Lee:

Yeah. Well, this has been a massive eye-opener, and my mind is blown in all sorts of directions. And I really, really appreciate you just taking some time and giving us all those insights. I don't even think I could summarize that entire episode. I think my biggest takeaway here is that AI will help us unleash creativity. Ai and machine learning and data altogether are wonderful, but you can't replace the humanity, you can't replace the strategy, and there are things that you have to be wary of, like reinforcing bad behavior. I think that's the best I could do to summarize this entire episode, but there is so much more in there. Folks, if you found value, please let us know in the comments below. Alex, before we say goodbye, what is the best way for people to connect with you? And then we shall say goodbye.

Alex:

Yeah, sure. They can email me at alex@clickmakers.io. It's clickmakers, not Masters.io, not. Com. I'm on LinkedIn, Twitter, Facebook. They can track me down.

Lee:

Folks, you'll be able to find all of those links in the show notes, so be sure to check out Alex, check out the website as well, clickmakers.io. I'm sure Alex will be able to help and support and answer any questions that you may have. Alex, mate, thank you so much for your time. We'll see you again soon, I hope.

Lee:

Cheerio.

Alex:

Thanks for having me.