DSP Strategy in the AI Era

DSP Strategy in the AI Era Speaker 2: Welcome back to the Better Advertising with BTR Media podcast. Today, I'm incredibly excited to have the one and only Gibby on the podcast. Now, you may not officially know him as Gibby. If you follow him on LinkedIn, you'll know him as Austin Hoggett. But let's start there. Gibby, why is this your preferred name? Why do we all call you Gibby? Why do clients now see you as Gibby within Slack? And when do we officially make the LinkedIn change? Speaker 1: So the name Gibby has actually been given to me by Adam, head of service delivery, and he and I actually go way back. We played college baseball together back in like the mid-20-teens, and essentially when I came to Pitt, my nickname was Biggin. Well, he didn't want to continue to call me the same thing that everybody else did, so he just flipped that around, added the BY to it, and that's how Gibby stuck. And ever since then, I don't think anybody has called me Austin or anything else besides Gibby. So that's how that has... Come to stick. And as far as the LinkedIn change, I do have it in little parentheses in between mine. So we'll see. We'll see when that goes through. Speaker 2: At some point, we need a full send it. We need to make this legal. We updated our Slack, which was a fun one as well, because I know, you know, when we bring on new team members or new clients, it always shows Austin and then all of us are referring to you as Gibby. If you are listening in and you see Gibby on the floor of Unboxed, which you will and probably more conferences in the future. Remember, it is not Austin, even if that's what you're used to seeing on LinkedIn. It is Gibby. Make sure to get that right. Amazing. Well, really excited for the topic today. Anyone who's followed BTR Media content over the last few months know that we've been talking a lot about AI, a lot about DSP, and a lot about AMC. Gibby, you lead a lot of that innovation within BTR Media. You head a lot of our largest spinning DSP accounts. You're doing a lot of fun things with AMC reports, but also AMC audiences. So I want to start off really high level What are the fun things you've been working on over the last few months? And then we'll get a little bit more into the weeds as well. Speaker 1: I think one of the most fun things that I've been doing is just understanding how to craft custom reports that help quantify all the strategies that we're running on DSP. That has been something that a lot of brands have come to us and Hey, you know, how can we understand how DSP is performing? How can we understand if it's actually driving incremental value? And so over the last few months, I've really been able to dig in, utilize AMC a lot more. And I've been having a ton of fun doing it because the information that you get back is just next level. And then as far as audiences, you know, audiences are the biggest thing on the DSP platform, utilizing AMC to its fullest capabilities, whether it's I'm leveraging your sponsored ads data to create an audience or just using it for basic ad to list or ad to cart retargeting. Sky's kind of the limit right now and it's been a lot of fun being able to utilize it over the last several months. Speaker 2: If you're listening in on this, Gibby's going to make you think this is easy. And I say that as someone who's done nothing but manage Amazon advertising campaigns for the last seven years. I managed a couple million dollars of AMG spend, which used to be DSP when I first got started. I have been doing this for a while. When AMC first rolled out, I would run on the treadmill and watch every single YouTube resource I could find on it, and it still overwhelmed me. And one of the main reasons it was so overwhelming is one, it's a black box of opportunity. It's not a perfect cookie cutter. Here's exactly what you need and I'm going to give it to you. It really takes someone who understands the business and can think creatively to come up with the appropriate queries to solve a specific business need for one. And for two, historically, you needed some level of understanding of SQL in order to really action on it, in my opinion. Well, one, Amazon's rolled out a lot of IQLs and no-code opportunities for the general audience. AMC is now available to everyone within an ad console. But they also, I think, have leaned into AI. You have leaned into AI and you've been using even just ChatGPT to help you with a lot of these areas. So I'm not going to discredit how much you know. You know the foundation way better than anyone else on the team, in my opinion, way better than me. But you're now using AI to act as a catalyst in terms of your AMC knowledge. Do you want to talk a little bit about that? Speaker 1: Yeah, so a little bit of background, like zero coding knowledge. Like I, you know, when COVID hit and everybody was starting to learn coding and basically you can figure out the code that says hello world on the blank page. I figured out how to do that. Now, stopped using it, got back into, you know, some other stuff, warehousing and all that, kind of forgot about coding. And then as AMC came up, you know, with the, with Better Media and everything, I wanted to learn how I could quantify, again, the DSP results. Some of the reports that are no code, you don't have to have the FSI subscription or anything through DSP or through AMC. Some of them do have some information that is pretty useful, but I wanted to dig deeper. I wanted to learn more about what my campaigns were doing, what the orders, what the line items, what the audiences were telling me. I just kind of opened a black box where I spent probably like three or four days just like, ChatGPT. Hey, I'm running this type of strategy. I'm running, you know, 90% of my budget upper funnel, 10% to lower funnel. I have a mix of STV, OLV, remarketing, competitor. What can I, what reports can I generate to quantify how all this is working, how this is building awareness and how this is actually driving incremental value for the brand. So it started with theorizing the question that I needed an answer to. When I started getting into that and started, all right, ChatGPT, help me craft a SQL code. Assume I have zero coding experience, which I do. Give me this, you know, give me a code that I can submit. I would submit it to the Amazon SQL or the MC SQL code. If it spit out a failure, I would, you know, copy the failure. Hey, ChatGPT, this was the failure. And then I would take a look at the schemas. The schemas are basically the little info on the side that you need to use for, you know, each parcel or whatever it's going to be on the SQL side of things. Take a screenshot of that. Here's the schemas that I believe we need to use. ChatGPT would update the code, and then I'd resubmit the code. So it was a lot of back and forth of, all right, let's submit. Let's see what fails. Resubmit the code. Let's see what fails. Let's see what I need to continue to fix. And as you continue to fix it, obviously ChatGPT continues to learn what your report needs to be. And as you update the schemas, ChatGPT understands the schemas more. And so then at the end of it, after, you know, two, three hours of back and forth fighting with ChatGPT on, hey, this is not right. Unknown Speaker: This is what I need. Speaker 1: It finally spits out the result. And that's where a lot of the reports, you know, the ones that I've sent just in general to you, to clients, that's where those are coming from. It's started with the theorization of what do I want my answer to be or what question do I need to get to this answer? How could I submit it to ChatGPT and how can I make it work for me? And then that's how I get to that point. Speaker 2: There's a few things I think are really incredible here. One, your output's only as good as your input. We see that with AI all the time. You have the understanding of Building the right context within your ChatGPT. And I think that's incredible because that's something that most people won't have. And most people are going to rush into this, go post their question to ChatGPT, get a terrible response and be done. I've seen it in all the WhatsApp groups and the masterminds of AI is just not working for me. Like, well, you have to take the time to build the foundation. You definitely need to train it and educate it. And that's where you are starting. You're saying, hey, you know, I didn't even start with like What do I need? You started with the question and what you needed and then you guided ChatGPT along that journey. So I think that's really important to call out for one. Anything else to add to that? Speaker 1: No, I mean, you just have to take the approach of you got to crawl before you can walk, you got to walk before you can run, you got to run before you can sprint. And so if you start with a sprinting question, obviously the input that you're going to get is going to suck. That's actually how I started and I got frustrated with it. Speaker 2: Terrible. Speaker 1: I would put it to me and I was like, okay, how can I think about this? This is a machine that needs to learn, you know, how did I learn? All right, let's ask the base level questions. This is what I want. This is what I need. Give me, you know, the output. All right, got the output. Let's see what needs to be fixed. What else do I need to run? And then when you get the results, the cool thing is, is you can submit the results and be like, hey, either explain this to me or hey, you know, these results don't look right. Here's actually what I'm pulling from the DSP. So there's like two different facets of all of this. You get your general report, you look at DSP. All right, hey, these numbers look a little bit inflated. Can you explain the delta between these and these? Speaker 2: I think that's the second most important part here is that education gap. That's something that I know is definitely an issue with brands when they're trying to understand the value of AMC is they'll look at a slide or a report and be like, how is this actionable? I mean, we were at a QBR together last week and three of the slides that were presented from a team that was not ours, we were all raising our hands of like, What do I do with this? It's cool seeing data on a slide, but if you don't know how to action on it, then it's really useless. It's just a shiny object. And I love that that's the approach you've taken for a lot of our brands is one, we do everything in a pretty customized fashion, right? We want to solve problems specific to our brand. So we sometimes will lead with that of, you know, what type of person are we presenting to? Is it someone that has had three espressos and is a little bit more analytical? Shout out to you, Matt, if you're listening. Is it someone that cares a little bit more about, you know, Brand awareness and brand lift, we're able to start from the person and start from the problem and then kind of work our way backwards and then use ChatGPT to help us educate that person as well. That's the part that I think is cool. I've screenshot or uploaded multiple decks into ChatGPT and be like, explain this to me like I'm 10. I actually did it yesterday. Alex sent me a voice note on taxes. And I copy and pasted the transcription to chat and said, explain this to me like I am 12. And now I feel like I'm very well versed in that topic. But that's an important part is how do you make it actionable? And how do you make it in a way that you can educate others on it as well? Speaker 1: Yeah, and that's actually one of my favorite, like, you just kind of hit it on the head, like, explain this to me like I'm five and 10. Because sometimes it'll water down to the easiest point of understanding to be like, You own a lemonade stand. You sell lemonade for $1. You sell five of them. You make five bucks. That's your end result. When you do that in tandem with, all right, hey, I have a solid prompt. That's the key that you pointed out too. You have to figure out the prompt. If you don't figure out the prompt, you're not going to get the results that you want. Speaker 2: Yeah. Some people try to rush through that part of it. It's like, that's the part where you really want to slow down, give it as much context as possible. And it's kind of cool to see if ChatGPT is already doing it. I think all of the retailers are going to build in this level of thinking. I mean, when AMC first rolled out, again, no code wasn't really a thing. It was super stressful having to figure out SQL. And then within two years of that original announcement, we've now escalated to the point of where we are now. Like the true future, in my opinion, of managing DSP, managing AMC and search and audiences is understanding a brand's overall needs. And then allowing the AIs to all do their thing. So we just have to prompt it. We have to be able to be the ones that are leading the strategy and say, hey, this is what I'm trying to solve. And then using ChatGPT to be the catalyst for that. Speaker 1: Absolutely. Yeah, I think you just hit it on the head, so. Speaker 2: Nothing much to add there. Out of curiosity, because I haven't played around with this much, I'm not sure if this is a dumb question or not. You had mentioned that your chat has learned so much more because of all of the prompting you've already done. Is that something that you feel like you need to redo every single time, or do you feel like your chat now versus then is much more educated on what you're trying to accomplish? Speaker 1: I will say there's always going to be that educational piece that you need to upload to the chat because sometimes the schemas in the back end of AMC do change. And I actually just ran into that problem the other day where I was like, okay, I've ran this code six times for six different brands. It's worked. And then a week later, there was actually one of the pieces of the schema that actually changed that's no longer available in AMC SQL. That I need to go through and figure out and change. And I was able to change it and submit it. And now, all right, hey, this is the new report. This is a new schema. So next time I go and upload the report or the code, ChatGPT will understand like, hey, this is a schema. And one of the big things that I've seen a lot of people do or that I've chatted with a lot of different partners or brands with when they're using ChatGPT is give it all the information that you can. So like one of the big things that I did, there's a couple hundred different schemas on the SQL side of things. Take it, copy and paste it, and submit it into ChatGPT. That way it has all of that information already. And you don't have to sit there and pick and choose what you think's going to be there. ChatGPT already has it by the different sections that it's in. Speaker 2: That makes sense. A big opportunity for us as well on the tech side to help API enable all of this. So that way it takes a lot of the weight off your plate. We'll have everything incorporated, documented from the get-go, and can pull it all in. Which is exciting. I mean, you take things like Amazon's most, one of their more recent partnerships on the AMC side was a partnership with Nielsen. And Nielsen is now opening up a lot of the insights that they have for AMC and DSP capabilities. So I think it was over 65,000 retailer insights, over 120 million household insights. You do have to pay a premium on the audience side. I think it was $2 view CPM premium, something along those lines. But Now you're not only targeting or pulling in analytics from Amazon, but also Nielsen. So you could get really excited. The majority of our brands are not just Amazon native. They're also retailer brands. We can now start piecing together, hey, when you spend this amount on Amazon, you also see a lift in XYZ. And obviously that's a step or a couple of steps that it's going to take for you to get there. If Amazon keeps increasing their partnerships with all of these networks and, you know, ESPN, Thursday Night Football, Disney, Netflix, I think it just provides more information that we have the opportunity to run with. Speaker 1: Yeah, and the amazing thing is, is as you continue to get all these signals from different partners, Netflix, Nielsen, you're going to gain access to all that customer data and where they're coming from, where they're going to, and audience creation, like the sky's going to be the limit. Whatever you can theorize around is going to be possible. And that's what's going to be really cool. It's like, all right, hey, all these different signals are coming into play. Now we can, you know, the fun really begins. Speaker 2: Fun, really. I mean, you combine that with Creative Studio, which has its bugs right now, but You can identify a very niche audience on a very specific platform and you can create that audience and target them with their own specific creative. And I think that's where like our roles are shifting is we're no longer going to have to do the heavy lifting that we did historically when it comes to bid budget optimization. Our role is now becoming much more creative. We're becoming more marketers than advertisers and that we're going to identify these audiences That are going to drive the most, the highest lift for our brands and then also come to the table if here's a specific creative that we know will resonate with this audience versus another audience. Speaker 1: Yeah, and that's the creative aspect in line with different audiences. It's going to be interesting. It's going to be a lot of fun. I mean, we're going to use like different creatives per step of the funnel and then just layering in all the different signals that we're going to be able to start pulling in. It's going to take it to the next level. Speaker 2: I love it. Well, this has been an incredible conversation. Do you have anything else to add for the audience or any pro tips that you would want to give? Speaker 1: Don't get frustrated. You know, when you submit your first prompt and you don't get the results that you want, just keep plugging away. The information that you're going to get is going to pay off. And like I said, I did all this without any coding experience outside of learning how to type hello world. Don't get frustrated. Just keep plugging along and The results are definitely worth it. Speaker 2: I love it. I think that's the perfect pro tip. It's progress over perfection in our industry. And if you're listening in and you want to learn more, Gibby is posting a ton of content on LinkedIn. And as always, we're always, you know, more than happy to answer questions and to help you along your journey. Both Gibby and I, and you know, the majority of our team had, you know, started from college in this industry. So we know what it's like to go through the learning curve that you're going through right now. So feel free to like, subscribe and share and we're more than happy to help with anything that we can. Thank you Gibby for hopping on and thank you for everyone for listening in. See ya.