Hello, and welcome to the Loop Marketing Podcast. I'm your host, Elise Stieferman, Director of Marketing at Coegi. Let's get started. Today, we are joined by Jake, Director of Data and Technology at Coegi as well as Candice, President of Radar Analytics. Today, we're talking about a very trendy topic, artificial intelligence. I know it's something our clients ask about on a regular basis about how we are using AI to activate, optimize, et cetera. But I think we need to start at the basics - talking about what is AI? And how does it really come into play for marketing activation?
Well, I think we should start off saying it’s a pretty misused term. AI is important. It's very present, but not necessarily everywhere. People think that they're leveraging AI. And I might not get things perfectly precise because I think to some degree there is an evolution in the definition of AI, but there are some kind of key concepts to it. Let's start with an algorithm. An algorithm’s like a set of rules where, “if my value’s above this threshold, do X, if it's below, do Y” so that's where a lot of computer-based decisions are made. If-then Boolean logic type of stuff. Whereas when you're introducing this concept of AI, there's not necessarily a rule written for the AI.
Those rules are determined based on some training set of data. And then it's going to take its learnings from that and make decisions on future data. When it makes these decisions, it's got a goal. You're telling it: achieve this goal and you're going to be able to change inputs. So it's going to change these five different things around to try to get to this goal, but it's going to make decisions, be able to choose what to prioritize, and do that based on the historic data. So, as it keeps growing and keeps learning, it's going to make better decisions because it's got a bigger set of information. “I tried these five inputs and it didn't work. I tried these five inputs and they did convert, they did buy my items so slowly.” These machines are able to hone in on what attributes are more and less important. And ultimately, it's the ability for some computer to make decisions and learn from those decisions and then keep making future decisions. So a lot of stuff is done in algorithms. Algorithms are very important, very cool, very related to AI - but there is a differentiation. And then even within AI, you've got machine learning and deep learning and all these other things, but they share some common characteristics where the computer is able to do human-like activities for us.
So then Candice, talk to me about the role of AI in marketing and data science today. What have you been seeing over the last five years in the proliferation of the conversation around AI and how do you think it's going to evolve thinking like the eventual cookieless future and other changes in marketing technology and just consumer life?
I mean, AI in our space is really just to help excel the ability to find insights quickly, to take out that need to dig for quick natural insights that can be found - trends, call outs, anomalies, things that aren't the normal - which is really helpful, because it allows the teams to have more time to do the digging for the bigger insights or the bigger business objectives. Instead of looking for those small month over month things that pop up that we didn't really anticipate or know that was going to be there. And it just gives you the chance to look at a bigger picture, which everybody wants, right? Everybody wants the ability to do the deeper diving than to do reporting. And that's really what we want to get at is that AI’s kind of taking care of the reporting side of things and allowing you to really get into that advanced analytics and deeper insights that you didn't really have the chance to before machine learning and artificial intelligence came into play as far as the cookieless future goes.
I think the main thing for everyone to understand is that your first party data is important, regardless of what's going on in the world, right? These platforms change on the daily. Now we're getting ready to have GA4 come into play and the methodology is very different from the original Google Analytics. So the main thing I think everyone needs to focus on is understanding the data you have. Figuring out if there is data you can bring into that story to really make it even more rich and using it and testing it and deploying it and identifying your consumer. The rest will kind of work itself out. You will always need a human. You know, a lot of times everyone thinks that these machines are going to take over the world. And they could, but you're still using that human aspect - that common sense, that ability to really process, from the beginning to the end. So I think just the importance of first party data would always be my take in that space.
And just add to that, data science is not necessarily a new field. Applying it to digital marketing and even just the speed at which you can do data science is definitely exponentially increased. But with these same technologies, it's like expected probability. How many times more likely is this user who's eligible to receive an ad right now going to convert? How much more likely are they to purchase my item than a generic person of the United States population? Some sort of like index level. So that sort of technology has, to some degree, always existed. The ability to activate on that in the microsecond ad buying environment is definitely changing, but that's when the concept of like cohorts or like groupings of people comes in. Where you can transact against this group with a certain expectation that they're more likely to do my desired action than just a generic user.
So I think the addressability might be going down. That's somewhat out of anyone in this room's control. That's just a macro environment we're playing in. So we're going to have to adjust the tools we have operating in a new macro environment. We have some agency to pick the strategies we're using obviously, but we're always going to, at the given time, make the best decisions on behalf of our advertisers as possible. So we're going to deploy the best strategies. We're going to track things as granularly as we can right now. The same technologies and the same approach, are going to exist in the cookieless future.
Our ability to track it down to who just bought my item might be diminished and there's a whole other conversation to be had about what's the right level. What level of personalization do people really like in their ads? But it exists both inside and outside digital marketing. It exists before and after the cookies, to some degree. It's just always in digital marketing. How fast can you adapt to the new environment, the new tactics available? So that's why we've got lots of different teams. So AI's important. It's a piece of our tool kit. But we need people that can, based on these decisions, go find more publisher-direct relationships. That's why it's not just me and Candice, there's whole other teams curating relationships and partnerships, and AI is going to be a piece of that, but there's a lot more of the cookieless future to consider.
I mean, it's going to be testing and learning what's important today. It's going to continue to be important to test and learn and reevaluate. And it's just a continual cycle between AI and humans. That's just part of our world. It changes so fast and so quick that you can't just set it and forget it. You have to sit there and you have to use that data in real time to make decisions and see if things are working - or if they're not, pivot from there.
So it sounds like, basically the human has to set the course and be able to allow the AI to optimize towards that course, but also have a bit of that human touch to understand the human experience, understanding the business that the machine can't understand. So when we're thinking about the application of AI and marketing, how much of it do we just hand off to the machines to do their magic versus what do marketers need to be looking for in that process to make sure that you're keeping in touch with the ultimate business goal?
Yeah, it's symbiotic. We want to work with the machines, not rely on them. There's different pieces of things that we do give it. And I think to some degree they make themselves apparent. So repeatable activities, things that you find yourself making the same sort of decision over and over - those are great. You know, when we're talking about AI and machine learning, those are the times where you apply it. For example, evaluating mobile web: how much do I value a mobile web impression versus a desktop impression? I can pull a billion rows in Excel. We'll pull millions and millions of rows in Excel, but you know, that's not where there's a lot of value in me deciding is it 1.5 or 1.4 and it's like a sine wave where you're going to eventually get to that right value where the machine learning is just going to be able to test all that out.
It's going to determine this, you said, “I want you to increase my return on ad spend as high as possible”. So then it's going to change all these parameters. And it's going to say, “when I use these parameters, I'm producing the best return on ad spend that I can for you”. That's great. I don't want to pull, you know, 10 decimal places out. That's definitely a spot where we can hand over the control.
When we're talking about, you know, there, there, and that that's full control, then there's some of these symbiotic, like I wanna work with you where we can run things like overlap reports where there might be a hundred thousand different audience segments, or a hundred thousand different publishers we can run on. And we can use the machines to say, give me a list of a thousand audiences I should use or a thousand publishers that I should run on.
So that's where you can use a machine to crunch a large amount of data. So there's kind of that middle area, but there's definitely a science, and our team is definitely experts, but there is a bit of an art to crafting a media plan. There's relationships where I know I've worked with very similar advertisers in the past, and we worked with these publishers because your audience is all on this publisher. So sometimes the machine is very shortsighted. It's like, okay, you're giving me a pick between these 10 ads to bid on. I wanna pick these three well. Well, it's worried about the next 0.1 seconds, where we've got a whole team of strategists where we're thinking three, five years from now, how does what we're doing today impact our overall favorability in the market? Which isn't going to have any return.
You're not going to see any ROI on that for three, five years. So that's where a good media plan is important. And chasing attribution can be very dangerous. So, not to go up on a ramp, but if you just target people who visited my checkout page in the last 10 minutes, but didn't quite make a purchase. You know, I could bid $200 on them, right? So just make sure I get that last impression and get attribution. So you might see your AI doing some real weird strategies like that, where it's not driving incrementality. Because that's the goal. I don't need to take credit for the purchase that was in process. You know, they were waiting for their credit card to process there on the next website. There's no way that you impacted that decision. So, how do you make a media plan that is targeted?
The right audience plan is going to fill the top of the funnel and not only do remarketing and paid search because that's the bottom of the funnel. That's the last step people are taking. You obviously want to own your search results. You have to be doing some of these things, but that's where you need a whole team of people to do that. So let the number crunching all be done by the computer and have our strategists work on making sure bigger picture things are happening. Identify when new opportunities are available that our media buyers are activating against those new strategies, right? That's the fun stuff. That's what they want to be doing. They don't need to be doing a lot of manual looking at the difference in, you know, two averages.
Well, and I think the other part about the human aspect is there’s still feelings behind your decisions, right? They're the brands that you decide to ultimately purchase with. There are feelings that can't be captured by AI, right? There are some people that like a certain dog photo, more than a different dog photo. That's not something that can be consistent across all people. So I think that's where you have to bring in that added level of understanding your audience and looking at it deeper at how people feel. That sentiment and understanding how you can compliment the sentiment and people's emotions with the metrics and information we're getting. I think AI will only continue to become stronger and more apparent the more the human can engage too and course correct and go in there and say, no, this isn't correct.
A lot of those platforms will allow you to say no, what you caught isn't quite right. Making sure that we are participating in what it's capturing will make it one day reliable to get to the point where it can even do a little bit more of that human and that machine aspect. So I think we can't forget the feelings. We can't forget the human interaction. We can't forget relationships and how they feel about you and make sure that's still part of the strategy that compliments the execution.
The one last thing I'd like to add is it has some freedom though, to kind of say, make sure things are in this tolerance range. We don't have to necessarily be watching things to a needle, but you look for the anomalies. You look for the ones that had a giant surge in spend yesterday. Why? Did something go wrong in the platform? You can adjust the job description of your existing staff to be focused on more inquisitive tasks. Like, why are things this way?
I jokingly and somewhat seriously say my job is reducing the number of mouse clicks it takes for every other department to do their job. Because if everyone just doesn't need to be doing little things, they're doing those higher level activities. So just preparing things, making sure that everything's going to, you know, put a bow on it, deliver it. so they can do what they're really good at. And it's, it's thinking it's, it's bringing in that context. Yep.
So I feel like AI is a very lofty topic for a lot of marketers. It feels inaccessible, it feels highly technical. So what can we do to make AI truly feel accessible to marketers at all levels?
Are there simple ways they can start to dip their toes into the AI water without having to have the Jakes of the world, with all the coding behind it, still be able to improve performance through AI.
Definitely, we all play a part. Whether you're upfront with new client onboarding or reporting out on the end, so all the way through the cycle - everyone has an impact. But a lot of it is just making sure you're critically thinking at every step of the way. If you have a plan to measure something, make sure that there are actual steps, milestones in the ground, on how we're going to get there. A lot of what people can control is giving good inputs. So making sure you've got good naming conventions. Accurate labels that are machine readable are super important, but good inputs also extend to your first party data.
So if you sit on a giant CRM of a hundred thousand people that have already purchased your item - let's go find more people like that. So if you find that data, start those conversations with any sort of legal process of making sure that it's done in a privacy safe manner. There's lots of different things that people can be doing, like a pixel strategy. So make sure you have a website that is conducive to a KPI. It's great that you've got all this content, but if you're after an email address, do you have a page that's optimized for capturing that email address? Making sure that you are doing things that are conducive to saying in a future state, I want to run a paid media campaign that's going to optimize to this one thing.
So I need some volume on that website. I need a website that is optimized for a good user experience because there's a lot of wheels that need to be greased along the way for when you're really ready to apply machine learning. Making sure when you're ready to apply all these things that you're not going to have some weak link in your process, because it really can't be broken. If you can't save your new leads into a CRM system to automatically email them, you know, send a follow up email to ultimately drive that person back to the website and make a purchase - this full consumer journey, marketing technology, AI - it requires a full circle that receives that feedback because AI needs to have a goal and it needs to be told if it's doing good or not.
So you've got to set up this plan. There are definitely some accessible AI in many of the DSPs we use. So even if you're not seeing a big flashing “I'm using AI” sign, there's all sorts of bid modifications that are being done based on. For example, I think that this person's likely going to be more likely to do my actions. So there's a lot of easy ways into AI with just running media with us. So we'll be employing some degree of it to optimize our campaigns. But everyone can make sure that they're thinking about data in a structured manner - enforcing naming conventions and auditing things. We run a lot of automated audits. Monitoring - are we within our contract guidelines? Making sure that we're set up to deliver successfully, because AI is just one piece of a bigger puzzle. So optimize everything else, and control what you can control
I agree. I mean, the foundation is the most important part. There's no point in trying to capture a lot of data. If we don't know the purpose or we don't have a bigger goal, doing an audit is a great idea. Even with the data that they currently have before we start using that first party data let's really dig into it. Is it, does it make sense as it clean? Is there something else that we should be doing with it, understanding what all they have available? Does it make sense to continue investing there or could we invest somewhere else that could bring even more value when you start implementing campaigns or executing and using artificial intelligence, that foundation is null and void. Spend that time up front, going through the discovery, creating that strategy, cleaning everything up, getting it prepared and then moving forward. So the only other thing I would say is don't be afraid to ask questions, find people that are experts in this space, have really smart conversations, allow people to help you set that foundation and then run wild. Start learning and capturing it and moving forward from there.
I think to sum up everything you all said, we can say safely that AI is not an easy button. It does help with the nitty gritty tasks, but you still need to have smart strategies at the foundation. And you have to put in the legwork upfront to really see the benefits.
Well, this was an interesting conversation. Thank you both!
Thank you for listening. Coegi is an industry leading performance marketing agency based in the Midwest. We've learned a lot since our founding in 2014 and started The Loop Marketing Podcast to share some of our hot takes on marketing trends we're following, best practices, we've discovered and actionable tips for improving your digital strategy. We'll see you next time!