Last year, MLtwist had the pleasure of speaking with Justin Campbell of Future Fuzz – The Digital Marketing Podcast.
When you think of machine learning, you probably think of a network of cold, humming supercomputers grinding through trillions of bytes of data.
You may not think as much about what happens after that data is transformed into meaningful insights and how those insights are applied to the most pressing issues in real-world industries.
In 2023, ad tech – a nearly 600 billion dollar industry – is at the forefront of some of the most relevant topics in digital media and online business models. How do brands optimize their ad spend while doing so in a way that is coherent with their brand image and values? How can machine learning technology prevent a communications crisis before it ever happens?
Here, David Smith talks with Justin Campbell of Future Fuzz about how ad tech increasingly relies on data and artificial intelligence, and how AI impacts the teams that have traditionally been responsible for operations like campaign optimization and brand safety.
JC: Hello, everybody. Welcome to the next edition of Future Fuzz. I’m delighted to welcome David Smith onto the podcast today from MLtwist. David, you’re calling in from very warm and sunny France. How are you doing today?
DS: I’m really well, Justin. Thank you.
JC: Great to have you on. We’ve been trying for a while, haven’t we? Because you were in the States and then you went to Cannes, but I’m really glad we asked you to finally make time in your calendar. Where there’s a will, there’s a way. So, David, I’m really curious, you actually requested to be on the podcast, which is a really nice request.
You know, I’ve checked out your company and I’m very interested in what you do. Could you just give us a fairly detailed background on, you know, who you are and where you’ve come from and why you set up MLtwist?
DS: Yeah, absolutely. So I have a computer engineering degree. I ended up joining a company called Falk eSolutions. It was based out of Germany and I was a sales engineer. They got picked up by a company called DoubleClick. And then DoubleClick got picked up by Google, and I spent a lot of time at Google focused on data and data partnerships. Then I ended up moving on to MarketShare, which was a company focused on using data for predictive modeling on how marketing was going to perform, marketing spend.
Then we got picked up by NuStar and NuStar got picked up by Golden Gate Capital. Soon afterwards, I got picked up by TransUnion. During that time, I made a move to Oracle to work on business development over there. Shortly after, I started MLtwist. So, a lot of time spent on data, I would say data for advertising, and what I had noticed was that machine learning was maybe 10 years ago, sort of a buzzword. Like,, if somebody said they had AI, it wasn’t very credible, it was kind of like, red flags would go off. But in the last three years, this has actually changed a lot.
Ad tech really is starting to embrace machine learning. What I noticed was a big bottleneck for data scientists is getting the data ready for machine learning. It’s actually a massive bottleneck to training the models so that they can go off and do what they want to do. MLtwist is fully focused on making sure the data is ready for machine learning and that data scientists can focus on modeling.
JC: Great. So how long ago did you start MLtwist then?
DS: MLtwist was founded in January 2021, so we’re heading into our second year.
JC: Great. Congratulations. So pretty much a COVID lockdown startup, would you say?
DS: Absolutely. Yes.
JC: Yeah. I mean, it’s a fairly interesting niche and we can imagine that a lot of people, you’re right, say, We do AI and we’re experts at AI, but a lot of it’s fluff, right? Companies are not really doing it. Who would be your ideal target customer?
DS: Sure. So in one of my previous lives I’ve seen machine learning applied to, again, predictive analytics. I’ve seen machine learning applied to identity and trying to figure out if an email address is a bot or an actual person.
Machine learning can be applied to contextual analysis. When you are about to put an ad on a website just getting a quick understanding of what that website is about. Is there any sensitive content on there? And machine learning is also being applied to brand safety. Again, fairly contextual, but if there’s a YouTube video, is that YouTube video aligned with the brand’s needs?
Like, those are decisions that need to be made in milliseconds. Those are all parts of ad tech, which maybe five to ten years ago is what we would call rules based. Like, you would see a couple words and and there would be words that were on a no-go list. Based on that, you would make a decision.
But slang changes over time. Things change over time. Things that were politically insensitive 10 years ago or politically sensitive now and vice versa. Machine learning really is something that a lot of technology companies are working to adopt in order to make those decisions in an ever-evolving environment.
JC: Where do you think we’re at right now? Do you think we’re at the very beginning of the impacts that it can have on people’s businesses, that they use AI properly and actually effectively? Do you think we’re at the very beginning or it’s already started? What are your thoughts on that?
DS: I think we’re about two years in. Where we are, I’m seeing that the models are still… If you look at the way models work, it’s a decision process. The reason why we don’t all evolve to models right now, I mean, obviously there’s the fact that, well, you know, you, you’ve got to make a lot of changes in your company, but really, models aren’t perfect in the same way humans aren’t perfect, right?
But humans typically are a lot better at concepts that are maybe a little are taking models a little longer to get there. What we are seeing is that models are good enough for that first pass. Like, hey, here’s a thousand URLs. Here’s a thousand images. Here’s a million images. And you throw a model at it, and then a model will go through it and go, Hey, okay. I think this 10%, somebody needs to take a look at, I don’t know what to do with, and this 50%, I’m sure there’s like, nothing really looks very sensitive, and then there’s this like, 40% you know, it could be, could not be.
I don’t think models are replacing, they’re more augmenting. They’re trying to help us navigate mountains of data and figure out where to focus versus right now, we don’t have a great method to do that.
JC: I think you hit something there that’s really interesting. It’s the “mountains of data” part.
I think people very often don’t know where to start with all the data that they have in their hands. And making that digestible and actually easy to understand what impact it has on the businesses is really important, right? I think that’s where people need help.
DS: Absolutely, yeah. It doesn’t even have to be brand safety.
When I was at MarketShare, we were doing media mix modeling, our new start multitouch attribution, there is a ton of data. Right away, whether it’s a simple Excel formula or some crazy model there’s already an understanding that nobody’s going to look at every single line on that row.
What do you do, then? Right, how do you help, you know, it’s always that old adage of trying to help brand: Fifty percent of my spend is wasted; I just don’t know which half. How do you help advertisers best spend money on what’s really going to give them value and that involves making sure they’re not advertising on content that’s inappropriate for their brand or making sure that they’re being able to target based on contextual with the cookie going away?
There are so many pieces out there that are coming into play. Again, I think some people are thinking that machine learning is going to take a lot of jobs. Me personally, I’m just seeing it, like, there’s just so much data. People just need help and they’re underwater with how much data there is.
Machine learning is a step to fixing that. It’s a garbage in, garbage out world. And if you don’t make sure that your data is good enough for models, like really good, like 98%, 99 percent good, then, then you might as well not be going down that rabbit hole. So that’s where MLtwist is focused.
JC: Yeah, and I think we’ve been here before where in the past algorithms and what have you were helping to optimize campaigns towards certain goals. I mean, the classic example is, let’s say, a cost per acquisition or a cost per download. And before it was an extremely manual task. You used to have teams of campaign managers optimizing campaigns and checking them almost on an hourly or even, you know, sometimes a permanent basis, depending on the spend.
And then there was the concern that these algorithms would then take jobs, but we’ve not seen that. They just changed. Yes, you maybe let campaign systems optimize themselves, but to a certain degree, there’s still so much left to do, so I don’t think that ever happened. It seems to me that it’s going to be a useful tool in the long term.
DS: Yes, we’re seeing that as well. Right now, customers aren’t coming to us going, Oh, we need to let a bunch of people go. It’s, Our people are underwater and we really need somebody focused on the data because our people are incredibly good. They need a first line of defense or a first pass to help the people on their side who are incredibly well trained and incredibly knowledgeable figure out where to focus.
JC: Great. I’m very curious what the next two years hold. I have that firsthand experience of using AI tools which generate copy, for example. The first company I started testing out was called Headline. And Headline was on Product Hunt, I believe, about five years ago. I got a free lifetime subscription.
Now when I started using Headlime, I thought it was okay. But recently, you can definitely tell that there’s learning there and it’s definitely improved. It actually is really good to generate ad copy and copy for anything. And it’s been everywhere, right?
DS: I mean, the first adopters were Google and Facebook, of these machine learning algorithms. It’s actually fairly recently that advertisers are really able to start to leverage machine learning for helping them do things. Before, if you wanted brand safety, you’d go to Google and they would say, Oh, okay, well we’ve done that.
We have something to make sure that everything is brand safe. Whether or not you would believe them, at least they had something. The difference if you went to a smaller publisher is that those smaller publishers would then try to go, Well, you know, we can go through this tool or that tool.
There’s this organization called GARM that’s out there and they are very focused on standardizing, like, if somebody’s smoking a cigarette, where does that fall on their brand safety matrix? Any technology that can classify these things on the fly across companies is pretty big for advertisers because now they can actually start to think, Oh, wow, I can advertise on places that I have technology that’s got my back and making sure that I stay brand safe, which is something that even five years ago was something that was very difficult to really convince an advertiser of.
JC: Yeah, extremely difficult. I can imagine. Great. David, so you’re in France now, and you’ve been to Cannes. I’m really curious, how was your experience of Cannes? I think it was the first time you were there.
DS: Actually it’s my third time but first time as MLtwist. So the overall experience was good. Every year is different. There’s a lot of heart that goes into it. A lot of times you realize you’re not there to strike a deal or sign a contract. You’re there a lot of times to listen. What’s going on?
What’s top of mind for the brands, for the agencies? You’re also there to celebrate. Cannes Lions in its core is meant to celebrate what innovative advertising technology or ads themselves are doing in the world. It’s really got a good spirit. People are not necessarily there to talk ill of anyone.
You know, they’re there to listen, they’re there to learn. They’re there to chat about what they’re working on and what they’re passionate about. Overall, for me, getting to sit in on a lot of the panels and a lot of the different events that happened throughout that week, as every year I’ve been able to go, has been an opportunity to reset myself and think, okay, this is what we’re about, this is where we’re going and, and this is where we are on that line.
JC: Do you think Cannes is more like a festival for creativity and thinking more than anything else? Because I’ve been to my fair share. I’ve never been to Cannes, actually. I’ve always wanted to go, and then Covid came along and Cannes was canceled for a couple of years. I’ve been to DMEXCO a lot in Germany, I’ve been to several smaller ones around Europe, I’ve been to some in the UK, of course. I’m wary of conferences. Is Cannes different in any way?
DS: Yes, yes and no. So it’s different off the bat. I haven’t been to DMEXCO myself, but it’s complete, almost completely outdoors, right? People are walking from booth to booth or from stand to stand, or beach to beach. I don’t know if DMEXCO is outdoors, but Cannes is.
That’s one of the few conferences I’ve been to that does that. The first year I went I was very lucky in that I got a ticket. These tickets are like 3,500 euros a pop. They’re not inexpensive. As a sponsor, we had tickets.
I actually got to spend a lot of time in the main hall listening to the presentations, looking at the exhibits. What I love about Cannes is if you’re a student, you can go for free if you live in the city of Cannes. I think you might be able to go for free or very little so it really is meant, in spirit, to be about something bigger than making the next deal. Once you’ve done that and you’re there, you’re there to celebrate the creativity. A lot of the agencies are there, a lot of the brands are there. As an add on, a lot of other people who want to get in front of those companies want to be there as well.
So I don’t want to be naive; you know, obviously, there’s always this aspect. But at its core, Cannes Lions is magic. It may not necessarily be a magical experience for everyone who goes, especially people who are mainly focused on the networking part. And there’s nothing wrong with that. But I felt the way you did, where before going, I kind of assumed it was just rosé and just people hanging out and networking. I went my first time and I was still a little new so maybe that was an advantage in that I didn’t have my calendar fully booked up but at its heart, it is something to really celebrate creativity. Lately, advertising has gotten into a lot of social aspects where you can actually have tearjerker stories on some of the things that some of the brands are doing out there.
JC: Great, that sounds like a really good summary. It’s still on my bucket list. Also South by Southwest in Austin. It’s just also a bit of a dream. I’ve been to Austin before, but it wasn’t for South by Southwest. And I was like, well, if I was going to live in the States, this would be, this would be my town.
Let’s touch on something you mentioned just a moment ago. It’s about the, you know, the social impact companies have at the moment and, and how it sits on their agenda. You know, it’s a topic that comes up frequently. I guess we’re talking about climate, we’re talking about diversity, we’re talking about fairness. A company that I’ve worked with before has just recently been listed as a B Corp, which is a great achievement for them.
It’s a hot topic for advertisers, right? How do you feel like technology, ad tech, and data are sitting in that space?
DS: Yeah. For me it’s very humbling because I get really excited about ad tech and then you hear the CMOs get on the panels and I just realize I always feel disconnected.
Karen Walker, Intel CMO, I believe she was on a panel and said something along the lines that brands are now expected to get involved on social issues. It’s not just about choosing, you know, which side you’re on, or, what part of the story you decide is the one that the brand stands for, but now you need to get involved.
Lorraine Twohill, Google CMO, called out something very similar and said that marketers were essentially – I’m paraphrasing – navigating how to promote brands, goods, and services before and now they actually need to understand how that fits in with the social issues.
While machine learning does do a lot of interesting stuff, the more altruistic – there’s – well, let’s call it fake news – out there. A lot of brands don’t want to sponsor fake news or that type of content. So there’s technology out there to try to stop those brands from indirectly or sometimes directly sponsoring a kind of content that they do not stand for.
But still, I feel very disconnected in terms of some of the issues around these concepts. It’s humbling. I wish I had a better answer realistically. The best I can think of is, well, if we get the ad tech right, then that should free them up to be able to focus on the things that matter. Which is where, frankly, ad tech probably cannot help them.
JC: You said something, though, I think, where [ad tech] can be picked up and that machine learning, artificial intelligence can be used as a tool to then help in so many ways. Like trolling, for example, or hate speech, or fake news where you can say we’re going to help combat these social issues that have a very bad effect on people.
Maybe one example as well is there was quite a lot of coverage in the UK media that there were these very young people talking about suicide in certain situations, and I’m sure that there’s something there where technology can help us support. I do appreciate that brand managers and CMOs are very concerned, I believe, like messaging and imagery, right?
They want to be seen that they’re doing something About certain issues, however, then they get sort of lambasted for either greenwashing or just doing it for marketing effects. Wouldn’t you think that technology actually could be really useful there in many cases, though?
DS: Yeah, absolutely. The thing that’s kind of scary and exciting about AI is you think about all the things that it can do. And you can let your imagination go with it. The reality, I think, is that machine learning has a long way to go. Data, currently, I believe, is a bottleneck. But is it really ad tech’s world to get involved in things like bullying? I don’t know the answer.
I also don’t know where brands want to get involved. Is it on their own websites? Is it outside of their websites? Is it in partnership with the government, or with, you know, professional suicide prevention? There are so many [possibilities] that to me are not necessarily ad tech. It feels more of a social science or therapeutic type of company that may want to pick up on that. To the degree that I think there is, there are concepts that do fall in our world. Things like potentially fake news, creative or what’s happening on a website to help steer brands.
But overall, I personally look to partner. I’m good at data, but I rely on other expert companies to decide what is the right decision. I absolutely agree that data is the way to get there, but in terms of which way do we go I look to other companies that are very focused on those subjects.
JC: Yeah, yeah. There’s definitely, there’s definitely impacts that we had. I know that there was the alliance for, I think it was better advertising [Coalition for Better Ads]. There was a lot of annoyance for a while, especially about display advertising and the type of advertising that you saw online: lots of pop-ups really in your face, you know, covering content. I think that advertising alliance was set up to try and combat that. I’m assuming it’s been relatively successful because you see it a lot less. It is way better.
Publishers are way more conscious of what they’re showing and doing, and for a lot of publishers it wasn’t always about making as much revenue as possible from advertising. They’ve tidied up their act. So it’s a starting process, right? And if the public requests it and the public starts asking about it, then you know, the managing directors, boards of directors, the CMOs, they need to pick up on it.
DS: Absolutely. And I think it’s, it’s good to see.
JC: Yeah. On a personal level, I was pretty annoyed at the reaction by businesses when it came to the Ukrainian invasion, like they were, they were covering their logos to the color of Ukrainian flags and sort of taking it to a bit of extreme, right? I wrote an article on that and I thought, there is a way to support and there’s a way to be aware of social issues in marketing, but it shouldn’t be used as an opportunity to promote yourself, right? I mean, do you have any experiences or thoughts on that?
DS: So right away, I want to point out that I’m a data person.
When it comes to those social stances. I’m not quite sure which way, you know – what is the right thing? When do you get involved? How do you get involved? I think those were points that were being made at Cannes between both Karen and Lorraine, and I’m sure other people.
Those are very hard questions. If somebody puts a flag on their site, I actually am okay with that personally. I think that there are times when you’re looking at somebody who does something and you’re like, well, are you doing that because you want to sell? Are you doing that because you really believe? I think that’s actually what it comes down to. When I see people who have a little sticker, that’s done relatively, in my opinion, tactfully, like, hey, we’re not trying to sell to you, we just want to let you know, like, this is how we feel about the war, or this is how we feel about fake news, or whatever it is.
If it’s done in a way that helps me understand that they feel this way without trying to use it as something to sell me something, that’s usually something I probably appreciate. But if it’s done in a way that’s clearly like, Hey, we’re doing this because we want you to buy stuff, then that feels a little disingenuous.
JC: Yeah, well it’s good you come at it from a logical data point of view, because then it’s less of an emotional sort of branding point of view, right? If you mean it, then it’s fine, but if you’re trying to make a buck or trying to make a statement, then that’s not that’s not honest, is it?
It’s a really interesting topic of conversation. Personally, I think machine learning has a huge impact in so many cases for advertisers and brand safety. I’m very excited to see what MLtwist will be doing in the near future.
Just to wrap up, I do like to sometimes ask the question to people who feature on the podcast: David, if you had met little David when he was, let’s say, a young, a young kid, and you could give him a piece of advice, whether that’s business-based or, you know, sort of life-learning based, what sort of advice would you give little David?
DS: I think it’s, Stick with it. I was a bit of a nerd when I was growing up. Magic the Gathering card game, card games, video games, Dungeons and Dragons, that kind of stuff. Back when I was growing up I think it was, maybe less cool than it is today? I don’t know if I’ve imagined that or maybe it’s still uncool; I have no idea.
JC: No, it’s really cool now!
DS: That’s true. I feel if you do something that you enjoy and that other people like – either it doesn’t hurt other people or ideally other people enjoy doing that stuff with you even if it’s not super popular – just stick with it.
It’s something that I did, but it’s not something I did with the knowledge that it was going to work out okay. It’s just something I did because I really couldn’t do anything else. There’s a lot going on in the world right now. So, you know, telling people to believe in themselves. And as long as you’re not hurting anyone, they’re going to be okay. I think that’s a nice thing to do.
JC: For you though, I tell you, like I can tell you right now, collecting cards and whatever you want to collect in the world, no matter how dorky it is, it’s fine because it is cool. Have you checked out that app called Whatnum? I’m showing it on my screen, on my camera now. It’s amazing. It’s an app for collectors. It’s a dream.
DS: Seems to be really popular with sneaker collectors. I’m seeing it, yes. What is this?
JC: They have live auctions from people who collect things like sneakers and game cards and baseball cards and you name it, you know, it’s on there. So, it is cool. Very cool.
DS: Yeah. Yes. That’s pretty neat.
JC: I think that’s awesome. It’s funny – when you collect stuff, it’s always got a personal story on top of it, right? There’s a physical thing that’s somehow collected. I’m holding it up on the screen now. It’s a collection. There’s a story you had to go through to get it to get that extra value. There we go. It’s like showing it on the screen. It’s a small selection of very unusual vehicles. Unfortunately, I lost contact with my buddy in Ukraine while I was training. But there you go. A little insight into my collection.
David, thank you so much for being on the podcast. It was an absolute pleasure. We’ll hopefully see you again in the future. Keep us posted on the developments at MLtwist. Thank you very much.
DS: Absolutely, Justin. Thank you so much. It was an absolute pleasure to speak with you.