Are we on the brink of advertising becoming too smart for its own good, or is Deep Learning finally getting us closer to what customers actually want?
Agility requires us to constantly evaluate how technology like AI reshapes the relationships between brands and consumers—sometimes for better, sometimes for far more complex. The advertising landscape is shifting under our feet, with new rules, new tech, and frankly, a lot of new guesswork.
Today we’re going to talk about how Deep Learning and AI are impacting advertising effectiveness, personalization, and the future of advertising—with or without cookies.
To help me discuss this topic, I’d like to welcome Jaysen Gillespie, VP, Global Head of Analytics and Product Marketing at RTB House.
About Jaysen Gillespie
Jaysen is a Southern California analytics pro with 15+ years in tech leadership. Currently holding the position of VP, Global Head of Product Marketing and Analytics at RTB House, he turns data into insights that drive relevant decisions. He is an experienced speaker and content creator, simplifying complex ideas and making them easily consumable and applicable. For Jaysen, analytics isn’t just interesting—it’s essential.
Resources
RTB House: https://www.rtbhouse.com https://www.rtbhouse.com
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://www.teksystems.com/versionnextnow
Catch the future of e-commerce at eTail Boston, August 11-14, 2025. Register now: https://bit.ly/etailboston and use code PARTNER20 for 20% off for retailers and brands
Don’t Miss MAICON 2025, October 14-16 in Cleveland – the event bringing together the brights minds and leading voices in AI. Use Code AGILE150 for $150 off registration. Go here to register: https://bit.ly/agile150“
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://www.theagilebrand.show
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
Greg Kihlstrom (00:00)
Are we on the brink of advertising becoming too smart for its own good or is deep learning finally getting us closer to what customers actually want? Agility requires us to constantly evaluate how technology like AI reshapes the relationships between brands and consumers, sometimes for better, sometimes for far more complex. The advertising landscape is shifting under our feet with new rules, new tech, and frankly, a lot of new guesswork.
Today we’re going to talk about how deep learning and AI are impacting advertising effectiveness, personalization, and the future of advertising with or without cookies. To help me discuss this topic, I’d like to welcome Jaysen Gillespie, VP, Global Head of Analytics and Product Marketing at RTB House. Jaysen, welcome back to the show.
Jaysen Gillespie (00:42)
Well thank you Greg, it is great to be here and great to be with you again. Yeah, yeah.
Greg Kihlstrom (00:46)
Yeah, looking forward to it. Yeah, as I was saying before we before we started recording last time I saw you was in Boston at etail. So it looks like we’ll both be there again. So looking forward to that, too. But but yeah, let’s in the meantime, you know, for those that didn’t catch our last episode together, why don’t you give a little background on yourself and what you’re currently doing at RTB House?
Jaysen Gillespie (01:05)
Yeah. As you said, I’m our global head of analytics and product marketing. And that really means thinking about what are the needs that we see out there in the marketplace across all countries. could be the United States, could be France, Germany, our home market of Poland, and making sure that we’ve got an offering that maps to those needs. It’s kind of like bringing the market to our product while also bringing the product out to the market.
A lot of that involves data analysis and making sure that things are operating properly. So there’s a natural join, I think, between people that love analytics and are analytically oriented and people that love bringing a product to the market. And you’ll often see like analytics teams embedded within product teams and things of that nature. So that’s kind of what I’m working on now for RTB house. And as you can imagine, there’s a lot of change right now in marketing. And I think we’re going to talk about some of those issues.
Greg Kihlstrom (01:54)
Yeah, yeah, absolutely. And, know, I do want to start with I know last time we chatted, we did talk a little bit about, you know, deep learning and, know, there’s lots of terms being, you know, there’s obviously we bring up AI literally every show lately, at least for the last couple of years. Probably some are familiar with the term machine learning. But I want to just make sure everybody has a good definition of what you mean by deep learning and how it does differ from particularly machine learning, but other types of AI in general.
Jaysen Gillespie (02:23)
Great question. I like to frame it in terms of some concentric circles with the largest most all-encompassing actually being AI. Even though you hear AI all the time, it’s really the least specific and most generic term for computers that are doing something like humans would do. And that’s really been around for decades. There was a computer back, I think, in the 60s that could play checkers. And it was a huge sensation at the time. Sure, it was probably as large as
An office, but nonetheless, it could play a game of checkers. That’s artificial intelligence. Well, as machines got more powerful and as more data came along, then the next circle in, I would call machine learning, which relies on machines figuring out statistical relationships between variables that exist in a set of data. Your classic, Hey, I want to know who’s likely to not pay their credit card bill. So I’m going to analyze all the payment behavior and compare it to the actual payments, I’m going to analyze all the behavior about people and try to predict something. Classic predictive modeling, machine learning. The thing about it is though, it requires like very like kind of normalized data tables and certain characteristics of the data itself. And then deep learning is really the part of AI that mimics the way a human learns. And there’s a few fundamental differences there. First of all,
Nobody presents you every five minutes with a nice, regular rectangle of data and says, here’s the last five minutes of your life. You are constantly receiving a fire hose of information through all your senses. You’re seeing things, you’re hearing things, you’re feeling things, you’re understanding things that you read. And that goes on a continuous basis. And your brain is sort of sorting it in real time into, hey, these are some things I might need to know for the next 10 seconds.
And here’s other things that, are really important. And I want to try to make a long-term memory out of that, or at least in some sort of midterm memory. And that’s how deep learning works. And I think the big innovation of RTB house was to take that principle of deep learning, the fire hose of information, the lack of a need for like a data scientist to kind of pre-process it, the ability to handle missing values or weird values or values that are out of the range natively and apply it to solve marketing problems know, chief among them are things like, how do I, you know, get more people transacting on my side? Or how do I get more new customers or address sort of non-obvious users that are harder to convert? These are all ways that deep learning, this is just fundamentally different from machine learning tends to excel.
Greg Kihlstrom (04:55)
Yeah, to kind of to build on that, then, you know, what are the the ways that deep learning is, you know, distinctively powerful? mean, you mentioned some things, you know, kind of, I think, inferring like unstructured data and some things that, you know, to your point, are not specifically in tables or whatever. But like what what makes it so powerful in the context of online advertising?
Jaysen Gillespie (05:16)
I think that there’s a few things. First of all, with online advertising, the nature of that fire hose better reflects the way the online e-commerce environment works. If you have an e-commerce website, you don’t get like a thousand people every hour that come to your website. They come sort of randomly. And even within a day, you might have an hour where only 27 people come and you might have an hour where 270,000 people come. And so.
Sometimes it’s a flood, sometimes it’s a trickle. And that’s very hard for, you know, machine learning algorithms to deal with things that have that strong variance or very heterogeneous user behavior. Some users come, they’re just kicking the tires. They’re not actually going to buy anything. Other users come, they buy something immediately. Other users come and they want to spend 18 sessions learning about a product because it’s an engagement ring with very high consideration. And deep learning can sort of natively ingest those very different behaviors and start to sort them out with enough users over time. And so it’s very suited to e-commerce. And then the flip side of that is what about the marketing piece of it? And it’s really the same thing. Your users then come to your website and what do they do next? Well, they go off into the ether and they go all over the internet and all over apps. And it’s also very heterogeneous, the websites they visit, the apps they use, the way they engage with the open internet. And so deep learning can marry those two together and drive value in ways that are somewhat different from traditional approaches. It can drive value on kind of an attributed basis, like, I’m scored in Google analytics or some sort of post-click system, but it can also drive value on an incremental basis because it tends to understand like what happens if I don’t expose people to ads because I never found them say out on the open internet. And what happens if I do expose people to ads and that difference of course is incrementality.
And incrementality is a topic that machine learning struggles with quite a bit more because it simply doesn’t have that holistic view. It mostly focuses on what happens once I show an ad. It’s very good at driving like an attributed transaction, but deep learning excels at the true incremental transactions.
Greg Kihlstrom (07:22)
Yeah, yeah, got it. OK. And so how are you seeing that deep learning? Let’s maybe talk a little bit about engagement across channel. I mean, you touched on this a little bit, but just to kind of go a little deeper there, you know, how does deep learning help brands engage more effectively? You know, things like personalization, relevance, even timing of ad delivery.
Jaysen Gillespie (07:45)
Yeah. So let’s talk about some of these, timing’s a great one because there’s a philosophy out there that if someone comes to a website and let’s say they don’t buy, I just want to like count them with ads everywhere they go. And that’s not really considering timing very much. And that’s because machine learning struggles with how to take timing into account. Whereas deep learning has the timestamp of everything, whether it’s on the e-commerce site or on the internet as a native part of that fire hose. And so it tends to be able to handle time related effects in a more sophisticated fashion. And one thing that we’ve noticed is that with deep learning, you often don’t serve as many ads sort of right away in like the first 15 or 20 minutes after someone’s left a website. Even though those ads will click pretty well and show attributed conversions, they may not be as incremental. And deep learning understands that and it’s a great difference and how timing effects work. The other thing, because it’s really different in kind, it’s just a different way of figuring out like who you should show ads to and what you should show in them. It tends to work very synergistically with other machine learning approaches. Most marketers nowadays realize, okay, I have a business objective, which is people came to my website and didn’t buy something. I want them to buy something. That’s not crazy. I think we all want that. What’s the right way to do that?
Well, the right way to do that is to apply all your resources, apply deep learning, but also apply machine learning or some business rules in a stack of related solutions that work together synergistically instead of competitively. And frankly, because of some changes in the SSPs where everything is now a first price auction instead of second price, those synergies are more self-evident, but that’s one reason why deep learning has become a powerful sort of powerful arrow in the quiver of marketers because they can pull it out and use it to supplement a lot of what they’re doing.
Greg Kihlstrom (09:39)
So can you talk a little bit and maybe share an example of, know, how how deep learning has helped improve campaign performance? mean, you know, certainly touched touched on that, but, you know, either in ways that maybe the marketers expected or perhaps in a way they weren’t expecting it.
Jaysen Gillespie (09:57)
Yeah. And I’ll probably give you one of each. So a lot of marketers, you know, their, their tune has changed quite a bit. Like I said, regarding building a stack of related solutions and five years ago, they would have really questioned me about that. But now I think the tide has changed and most marketers have an expectation that if I bring in deep learning, I’m going to get more, I’m going to get more attributed sales. I’m going to get more incremental sales at sort of the same row as, and that seems to play out. And it’s nice to kind of confirm what marketers think about that on the flip side.
Sometimes it’s nice to present something that’s maybe different from what marketers are used to. like to call it like a marketing myth buster. And there’s a myth out there that, well, people that have bought from me, or maybe they bought from me twice or three times or some criteria, I don’t need to market to those people anymore because I’ve already established such a relationship. And we’ve done some studies, some incrementality studies where we carve off like a carefully constructed randomized control group.
And not shown ads to those people. And then also shown ads to those people and actually shown that at least with deep learning, there’s a lot of incremental gain to continuing to market to those prior buyers. And when we ask prior buyers, you know, what we basically are is, Hey, I bought something from retailer X and it was great. I got good shipping. I got a price. liked the product itself was good. There’s a halo effect there that actually greases the wheels for future marketing. It actually makes it easier to drive an incremental transaction from someone who’s already bought from you and had a great experience. And a lot of times that’s a real aha moment for marketers who have sometimes like created, you know, holdout rules or exclusion rules, and they’re able to undo those when they’re given, you know, some evidence and a reason to believe, well, I know you should actually keep marketing to those people. And in fact, it’s going to grow your top line, you know, fairly quickly.
Greg Kihlstrom (11:50)
Yeah, yeah. And so I guess, you know, to build on to that, so, you know, that’s a great example of AI kind of help teaching, maybe breaking down some assumptions. Again, there’s reasons for those holdout rules and and all those things. And yet at the same time, AI is able to help us understand that doesn’t always apply and it’s the right thing at the right time for the right person and all that other stuff.
What’s the role of humans in this, you know, with, with, you know, with stacks of AI, know, with machine learning, deep learning, you know, all of these things and, all of these things helping us, you know, with the next best offer action, all that stuff, you know, what, what, what’s the role that humans still play?
Jaysen Gillespie (12:28)
So I think you need to look across the spectrum of jobs that AI might affect influence or even completely do. And there’s a lot of jobs out there, hopefully not an enormous number, but certainly some where you’re just looking at people literally like cutting and pasting things into Excel sheets and G sheets. And you’re like, this job, this shouldn’t be a job. This, this should have been a job that went away 10 years ago. And I think those are the things where AI is finally going to be like, all right, we can do this in an automated fashion. It was just never worth the automation effort because the cost of coding was more than just keep letting, you know, Jim over in the corner, run your numbers every day. Well, now you can just, you know, automate that by having the AI write the code or using an AI agent that can handle spreadsheets or even natively in G sheets. For example, it’s got a lot of cool like data analysis. You can talk to it and say, Hey, analyze this data for me. Wonderful. Thank you for that. So that most of this stuff is just pieces of people’s job, right? But there are a few jobs that shouldn’t have existed for years. And those are like the low hanging fruit to say, Hey, let’s bring in an AI to do this. And there’s a good chunk of that in marketing, especially at agencies. And that’s why you see like layoffs and stuff like that at an agency where media planner, for example, like that’s a very old school, like I’m juggling numbers in spreadsheets and probably won’t exist.
Yeah. If it even still exists today, on the flip side, you know, AI is often wrong in that, some sense, it’s very human. It’s not something that you just want to kind of let go on its own. And it needs a lot of oversight. And sometimes it needs, you know, re steering it and things of that nature. And that’s where humans have a real strength is we come in and we look at things with a critical eye and we review and we say, does this make sense? Is this right? Do I want to believe everything that company X, Y, or Z is telling me. And so there’s still a huge opportunity for humans at a slightly higher level, I think in marketing in a number of places. I think you’re also going to see, you know, you’re hearing a lot about AI and creative because it can simply generate, you know, 4,000 versions of creative. But if, if you’re hearing 4,000 versions of creative that aren’t that good versus two that really are.
Greg Kihlstrom (14:29)
Yeah.
Jaysen Gillespie (14:43)
You know, then that’s a great AB test is, is this AI really helping me? So I think there’s, there’s always going to be the opportunity for some AI and then opportunity for a human touch, and a human who really understands, you know, that, I don’t need like another shade of red on the sweater. What I actually needed was the sweater, you know, flying around the empire state building for whatever reason. And I may not think of that because it’s so novel and off the chart compared to what the AI was trained on. AI is going to be big in making marketing work better, especially on like the publisher side. A great example, we’ve rolled out a technology called Intent GPT, which basically is the AI version of contextual targeting. You’re probably familiar with like contextual targeting, which is where people say, hey, I want to show ads for sweaters. Anytime there’s a web page with the word sweater, I want to be on that page.
Well, that works about as poorly as it sounds, right? There’s a guy that’s choked another guy to death using a red sweater. You don’t want to be there, but you’re going to be there using contextual targeting one. Contextual targeting V2 is the AI actually reads the internet, reads every page and reads every word and says, wait a minute, this is a gruesome story about a brutal murder. Let’s not put your ad on that. Let’s put your ad over here where it talks about, you know,
There’s a cold front coming and why you need to dress appropriately and things that would be much more useful for a sweater intender or, or indicative of someone who would be interested in your products. So AI is going to hopefully supercharge some ability to be much more specific on the publisher side. It’s going to have some good impact on like the creative and the personalization that kind of we’ve known and loved with machine learning, deep learning. We’ll, we’ll come in and take that to the next level. And then it’s going to be your assistant with a lot of human oversight, I think on the job and personnel side.
Greg Kihlstrom (16:35)
Yeah, yeah, that makes sense. And yeah, I mean, you know, a couple of things you mentioned, like on the creative side, like as well as on reviewing all of the potential. So I mean, that’s a scale thing that humans just aren’t great at. You know, like that’s that’s a great use of of a I initial creative ideas. Usually better coming from a human. But again, doing that in all the formats and all the ways and all that stuff like that’s great for a I and know like that’s right. People don’t really want to make
I’ve made multiple versions of banner ads back in the day. Like, no, nobody really wants to do that. It’s a job. But AI does that so well that, you know, it pushes the humans to do better work and more valuable work and stuff. So, yeah, I think I think that’s a great way of of couching that. One other thing I want to talk about before before we were up here is just, you know, cookies and, know, there’s been a lot of back and forth and back and forth on this over the I forget where we were last time we talked as far as with
in the world of cookies. But where we are today, at least, you know, it sounds like they’re not really third party cookies are not really being deprecated. Who knows? Probably the day after we record this, some some other change will come out or whatever. But I know. But at the very least, you know, what does this mean? Like what what should advertisers be doing right now in this world where, you know, at least for now, they’re not being deprecated. But we still know that.
Jaysen Gillespie (17:44)
That’s the fun of it.
Greg Kihlstrom (17:57)
You know, there’s a lot of benefits of first party data and all that, you where should advertisers heads be outright?
Jaysen Gillespie (18:02)
Yeah, you just nailed it. So you always want to be on first party data and the third party cookie, you know, is it, will it, won’t it, that whole drama, the big positive of that, I think are two things. One, it trained people that you need to prepare for a future where addressability is not going to be universal for any cohort. Some people may have a third party cookie.
Some people may be on other browsers that don’t have a third party cookie. Some people allowed you to share their IDFA, for example, on app. Some people are accessible via a third party ID of some sort or a publisher ID. You need to meet people where they are. And you need to meet tech where it is. And you’re going to need to have partners and technical solutions that handle all of those use cases as best you can. So it’s taught us that I think the industry has learned that.
And Google sort of had to come back to this thing because I think they were in danger of just completely losing control of identity because while they were deciding whether or not, and how they wanted to handle the third party cookie, a pretty robust alternative identity ecosystem has started to spring up. And you know, that’s got to be yet another thing that scares them to death every night because it’s yet another point of control that they’re losing in terms of like Chrome ownership in their case.
Greg Kihlstrom (19:21)
Yeah.
Jaysen Gillespie (19:22)
It’s fundamentally good though, that Google has finally put a stake down because one it’s resolved in uncertainty too. It’s trained everybody that we’re going to be in a multi identity universe. Three, it’s allowed a neat ecosystem of other options to make themselves available. And four, it’s taught marketers that, know what? It’s all about your first party data anyway. And the less you rely on these third party, you know, data ads, which are usually wrong. And the more you dig in.
on what you know for certain, the better you’re going to do. And you want to leverage first party data. You want to do what we call first party advertising, which is do more with your first party data instead of doing less and relying on, well, I’m buying an audience from this sort of murky black boxy audience vending machine where I put a quarter in and I get the little plastic thing. And I don’t know what toys inside there. Maybe it’s something I want. Maybe it’s something I don’t want.
Greg Kihlstrom (20:19)
Right? Yeah, no, I love it. Well, Jaysen, always great to talk with you. Thanks. Thanks for your ideas and insights here. One last question for you before we wrap up. What do do to stay agile in your role and how do you find a way to do it consistently?
Jaysen Gillespie (20:33)
You know, it was funny. I was asked a similar question. I said, you know, the best way to stay agile in your role, I would argue has nothing to do with your role. Agility is a mindset and you build agility by doing a lot of different things in my view. So it’s like, I I’m a notoriously fast Walker around New York city. I learned Polish for work. You know, I read, you know, a variety of unusual books. You know, I attempt to play some golf.
These are all agility building things, even though they’re not related to ad tech or what you’re doing. Yeah, of course you’ve got to stay up on the latest trends. And I wouldn’t ignore, you know, the trade press for a full year or something like that. But I’d encourage people to flex their mind in ways outside of your formal role. And sometimes you find like on that fast walk around New York city or Santa Monica, you have some moment of insight because you were away from work and not specifically thinking. And I think that leads to agility.
