← Next in Media
Next in Media
Amazon's Kelly MacLean on How the eCommerce Giant Built a Killer DSP
00:21:28
Next in Media talked to Kelly MacLean, Vice President of Amazon DSP, about her journey from professional soccer to leading one of the industry's most talked-about demand-side platforms. MacLean discussed Amazon DSP's evolution from a single-purpose tool to a comprehensive multi-purpose platform, the technical rebuilding required to compete at scale, and how AI is transforming campaign automation while maintaining advertiser control. The conversation covered Amazon's approach to streaming TV inventory, their partnerships with major publishers like Roku and Disney, and how changing consumer behaviors are shaping the future of digital advertising. MacLean also teased upcoming innovations to be revealed at Amazon Unboxed in November.
Loading summary
Transcript37 lines
- [00:00]
A
Freemantle, one of the world's most prolific creators and distributors of premium content such as America's Got Talent, the Price Is Right and Family feud, reaching over 100 million US households, leverages Elemental TV's AI powered one audience platform to offer advertisers access to its extensive library of iconic titles all through a leaner optimized supply path. Lesser intermediaries enabling increased brand safety and more cost effective campaigns. Discover how@ elemental tv.com One audience foreign this week on Nexting Media, I spoke with Kelly McLean. She's the vice president of possibly the hottest, maybe most feared platform in ad tech, the Amazon DSP. McLean and I talked about her move from Meta to Amazon a few years ago and our team has tried to evolve a tool that was once built mostly for small brands running sponsored ads on Amazon.com to becoming a major player in TV, the open web and working for all sorts of advertisers. So much so that she and Amazon are making the trade desk nervous. Lots to talk about. So let's get started. Hey Kelly, thanks for being here.
- [01:03]
B
Hey Mike, thank you for having me. Excited to be joining you today.
- [01:07]
A
Same a very exciting time to talk to you as Amazon DSP is getting just a lot of attention excitement. Why don't we back up a little bit though? Can you maybe just you came, tell us how you got into the business, give us a little bit about your background. Who are you?
- [01:19]
B
Yeah, definitely. So actually my first career was playing national team and pro level soccer. So funny enough, I actually never dreamt I, you know, end up in ad tech. But you know, ultimately at my core, you know, I've always been very competitive. I've always loved moving fast, building teams and I've always enjoyed technology and was, you know, really fascinated at the rate of change, inspired, you know, early on by Moore's law, you know, driven by system level design and novel computing programs. And I actually saw this firsthand at my first job which was at Dell and I got awesome exposure to manufacturing, sales, product, lived and ran a team in Central America and realized I, you know, I really wanted to be at the core of building products and became fascinated with startups in Silicon Valley. Went to Stanford for grad school and ended up working in product for a couple of startups in the Valley, managing large data graphs and real time display video ads. If you remember when that was kind of the hot thing. But you know, I think ultimately I just, I wanted to learn at scale so I moved to Facebook at the time and you know, was fortunate to build and Work on all parts of the ad tech stack, help build their off Facebook mobile network brand direct response verticals, video, the early app install ad helped to build out early emerging markets businesses and moved to London to open their first ads engineering office and then most recently kind of a more central engineering org, managing the entire customer graph, scaling multiple small business products across the family of apps. So it was a fun journey moving from roles in marketing to product. I taught myself how to code. I took on leading large scale engineering teams and it was just a really fun time taking the business from about 4 billion to 100 billion over that time.
- [03:03]
A
So you were not on the pitch like just dreaming about the way you're going to use data and computing power over time. This was not the dream of the soccer field necessarily.
- [03:11]
B
Exactly. Yeah. It's a different pitch. I moved from one pitch to a different pitch.
- [03:16]
A
Okay, so a couple years ago you make the move, you go from Meta to Amazon. What's behind that? You're at a company, you're going from an ad juggernaut to one that is like taking off what was and changing what was behind that move. You're going from an ad juggernaut to one that is like taking off and changing what was behind that move.
- [03:32]
B
Yeah, really good question. You know, again, I've been fortunate to work on all parts of ads. While I appreciated the strategic focus on the metaverse, I didn't have a personal connection to that direction. And I really loved ads. And yet as ads had kind of matured over time, I was just yearning again for that next big challenge. You know, I'm a builder at heart. I've always loved, you know, either zero to one opportunities, looking at greenfield areas, pulling together the roadmap and bringing together a group of scrappy engineers to execute and scale, or taking something that was more mature, re architecting the system and rebuilding the team so we can take it to the next level. And Amazon had actually reached out a couple of times before I made the jump. And I was really inspired by, you know, the vision to re accelerate the Amazon DSP for large scale growth. And Amazon's customer obsession really resonated with me. And at the time they just made the move to bring together all parts of their core engineering and science teams into a single org, which I would leave. And so this paired with just the trillions of real time purchase signals is an engineer's dream to track, to tackle. So I made the jump.
- [04:39]
A
So yeah, you go to, you come to Amazon. It's not exactly a green field maybe, but it's, it's not a rebuild either. Like the DSPS exists, but it's maybe not as prominently talked about as it is today. Give us like the history there of like, you know, I believe the DSP was really just built for Amazon and then it's become more than that. What was going on when you came in and what did you learn about the history?
- [04:56]
B
It's a really interesting question. I think if you think about Amazon dsp, you know, it really has evolved over more than a decade and started as more of a single purpose DSP serving our stores and growing businesses. This was kind of the pivot to come in was to really help make that shift from single purpose to multipurpose. You know, the digital landscape as you know, it's rapidly evolving. Streaming TV is becoming mainstream, full funnel is becoming more critical. Cross channel reach and measurement are starting to take center stage. And so I think it was really about, you know, a couple of years ago and with me coming in, we had to pioneer ad tech. And this was a really opportunity to just make it easier for all advertisers to navigate this evolution with precise reach, deeper insights, improving advertiser outcomes. And so I worked with the team to make sure that we were building a system that would drive growth for all digital marketers across verticals and advertiser sizes and really kind of pivoting our focus, leveraging all of the power of what had been built and the power of the integration with the Amazon store, but thinking about what are the investments we need to make technically to help us scale, you know, for the next five to 10 years.
- [06:05]
A
So on that, you know, I remember a decade or more ago when a lot of DSPs and ad tech companies that were built for the display era all of a sudden have to figure out how to pivot into video. Amazon was, the DSP was, it wasn't as though video was new, but all of a sudden in the last couple years there's just a whole lot more streaming ad inventory. How did you have to adjust there? What did you have to build or not build in terms of that big shift?
- [06:26]
B
So I think, you know, our goal for Amazon DSP is really about providing choice and flexibility to drive those meaningful moments between brands and consumers. And so, you know, we want to make sure that brands can find customers, find where they're spending their time. So we wanted to make sure that the Amazon DSP is the best place to buy advertising while consistently inventing ways to increase efficiency and lower the cost of managing campaigns. But all backed by AI and so this also means for marketers it has to be easy, right, to deliver across all premium publishers. And that's why we've continued to build strong partnerships with publishers as well as other partners in the ecosystem like Roku, Disney and others, and making more and more significant investments to invent a world class suite of interoperable ad tech solutions. And we're seeing really positive response from media buyers, brands in the industry, which we think really makes us strategically positioned to continue to deliver those full funnel outcomes, you know, across all premium media.
- [07:26]
A
So lately there's been like, obviously there's been a lot of excitement around Amazon DSP and we love these like creating these wars, right, the warring DSPs and who's going to win. I always say to people, I've covered this industry a long time, but I don't use this stuff like you would. I don't have hands on experience in DSPs, but the perception is, well, they're probably, are they the same and that, but you know, and that like one has better inventory or better data than the other. How do you kind of articulate like the big technical differences besides just we've got awesome data and awesome Amazon ad inventory. Is that the way you think about it or how do you explain that?
- [07:54]
B
Yeah. And Mike, we're gonna have to get you a demo and we're gonna have to get you using Amazon. Yeah, exactly. We've simplified, we've made it really cool. Yes, I wouldn't expect you to. Yeah, I think that's, that's, that'd be frightening a little bit. Yes, exactly. Yeah. So I mean, I think taking a little bit of a step back on, you know, our vision and kind of how I think about as a builder, you know, our approach for Amazon DSP and then I think that kind of like ladders into your question a little bit more directly. You know, I think when initially looking at our systems, we recognize that we're going to have to really re architect Amazon DSP for scale, as I shared earlier. And so our vision has been to really work with media buyers and partners to be the most performant AI buying system for all digital marketers of all sizes. And so in the last couple of years, just to kind of bring you on the journey a little bit, we completely rearranged, architected the back and front end, moved from that single purpose to fully functioning DSP that's capable of delivering all of our signals across the open Internet. And again all backed by machine learning and AI. And this was really critical. Right. We had to rebuild Our entire ad sourcing system, our models, we had to layer in the AI into every step of how we were building and delivering our products. And this really improved overall performance by over 40%. And so I think, you know, to answer kind of your question a little bit more directly, you know, DSPs can be complex and historically this was a feature, not a bug, right?
- [09:20]
A
Yeah.
- [09:20]
B
And yet kept adding stuff to it.
- [09:22]
A
Making it more complicated over time.
- [09:25]
B
Yeah, absolutely. And yet customers have been telling us they want simplicity. So we're making significant strides in simplifying and automating Amazon dsp. And additionally we want to be the preferred DSP and make it easy to directly connect advertisers and publishers so that they're delivering the most relevant messages to consumers. We're continuously expanding our reach across the open Internet and making it easier for publishers with the power of our first party. Streaming, browsing and buying signals through privacy, safe clean rooms. So ultimately, you know, we want to make sure that it's an easy way to connect pubs and advertisers. We want to make sure that our pricing has been very simple. We've been very thoughtful on pricing, so we've been helping advertisers save with industry leading low fees. 0% for programmatic guaranteed deals on Amazon properties and 1% for programmatic guaranteed deals.
- [10:15]
A
Sounds good to me.
- [10:16]
B
Yeah. Across premium streaming publishers. And so if you boil all of this down, you know, our key differentiators are really those signal the trillions of buy, browse and streaming signals, the premium inventory and reach. You know, we have the largest premium reach across the open Internet paired with deterministic identity across Prime Video in the majority of all premium pubs, which as we announced it at can on our partnership with Roku, we're now the only DSP that is offering authenticated reach to over 80 million CTV households in the US and then ultimately just kind of underpinning all of that with tech and interoperability. So all of the investments that I shared just make it so that we're much more interoperable with existing Amazon products, with third party partner products and measurement integrations. And we want to continue to be agnostic and bring hubs and advertisers together via Amazon Marketing Cloud, Amazon Publisher cloud, our clean room suite as I shared, and then just better enabling partners around the ad tech ecosystem.
- [11:17]
A
So you talked about all the rebuilding you've done and you're trying to bring this out to the open web and a broader array of partners. Is it hard to build something that I imagine you're trying to serve like the core. Maybe the traditional Amazon advertiser that has worked with you for a long time has tons of data and learning versus brands that are doing more, CTV versus brands that are not selling on Amazon necessarily. Like, is that, how does that evolution look like? Is that hard to build a product that works for everybody like that?
- [11:45]
B
Well, I think the nice thing is because you can leverage our signals across Amazon and across the open Internet. And we're really bullish about how our signals can drive marketers, businesses. And so there was, you know, areas that we needed to beef up and accelerate, as I shared earlier, because we just felt like for being a full multipurpose dsp, we just needed to make sure that we had all of those features for marketers to truly scale their business, whether they were longtime Amazon customers selling products on Amazon or they had a tangential service, right, that, that they hadn't been selling on Amazon. So I think for us though, it was how do we work deeply with traders and with brands and agenc to really understand where are those pain points, what are the outcomes that they need to drive their business and then how do we make that more seamless? And ultimately we want advertisers to have more working media to drive outcomes for their business. And that's been a key part of our offering is providing more transparent tools that effectively put advertisers in control of their investments. You know, and that's something that we've been really invested in as we're expanding our reach across the open Internet, enabling advertisers and publishers with the power of our signals. You know, for example, as I shared at can, we announced integrations with Roku and Disney and you know, across Amazon's O and O properties, we and premium publishers, we believe that advertisers ultimately executing their campaigns on our DSP is going to drive and has shown to drive better full funnel outcomes and more efficient spend. So it's really about bringing together, you know, that precision and that performance to just continue to create those beneficial experiences, which is really core to what we do at Amazon.
- [13:25]
A
So you mentioned a couple of things which, which might sound like they might be contradictory in this new world where you have brands definitely want to have like the smartest technology on their side in terms of optimization, you know, AI driven media buying. But then they also want, they also sort of like to have control, which is a long way of me asking like what happens with this product over time as more automation and AI driven media buying comes to be the norm or seems to be where we're headed.
- [13:48]
B
Yeah, absolutely. I mean, AI is, you know, so transformative and it is driving change, as we all can see, at an accelerated pace that, you know, I don't think we've ever experienced experienced before, which is fun. And I think our approach has been that we really want to provide more of a crystal box. Everyone talks about the sphere and these black boxes, right, that you put some inputs into an AI or some prompts into an AI system, something happened, see what happens. Right, exactly. And our approach has been let's offer a crystal box, right? Let's make sure that it can be easy. And if you want to set up with a couple inputs or prompts or constraints, you can do that. But we will still provide you the transparency, optimization, control and transparency on reporting so you know where your ads are going. And in addition, if you want kind of more expanded full featured systems, you can have access to that as well. All of that will be powered by AI as it is today. And so, you know, I think just given that technology is changing this consumer behavior and reducing that journey to discovery for brands, there's this need to shorten the time to build campaigns. And so that's been a big focus for us on, you know, building Performance plus and Brand plus, which, you know, these are industry first where we're combining that power of automation with that strategic over and that control as I was sharing, that gives advertisers, you know, more of that actionable capabilities to control their campaign, but also that transparent reporting while leveraging machine learning for data driven optimization. So you have more understanding over inventory management, brand safety. And we've seen performance plus driving over 51% improvement in customer acquisition costs. Brand is industry first in that we're actually trying to power leverage AI to power brand objectives. And again, this is really this kind of crystal box capability that we've introduced and then we've also come out with complete tv. This was, we launched this ahead of Upfronts this year, which leverages AI to help marketers plan, manage and measure more of their holistic streaming TV buys across Amazon and leading streaming apps to deliver spend commitments in full. And so again, this is a really neat way where we're leveraging AI to simplify campaign development as well as smarter decisioning for execution of scatter dollars across the open Internet. And all of this is underpinned by ad relevance, which is our innovative approach to kind of determine and find ad opportunities, leveraging AI. So we're seeing massive transformation in not only, I think the first kind of evolution with AI and ads, a lot of it was really centered around creative. We've been really investing in creative. We made a lot of improvements and we have some more fun things to be announcing there. The next phase is really about campaign development, automation, and then how you start to bring all of these pieces together. But we really want to do that with our partners and help to empower them with more of those controls so they can make even more smarter decisions.
- [16:48]
A
Is anybody actually, you know, there's a talk of a future where you'll just have an. You'll just throw your agent at the DSP and it'll do everything for you and no one will, no person will ever be involved. Is anybody actually doing that now or is that still like a theoretical thing if it ever comes to fruition?
- [17:02]
B
It's a good question. There's a lot of experimentation going on right across the board and agents will definitely be important part of ad development. I think the question will be, as you called out, right, what is kind of that degree of hands off the wheel that you will see in the building process. And it likely will be a combination of offering simplified products where you can leverage agents and some prompts and some constraints paired with. For certain campaigns, you're always going to also need the more complex and kind of expert features. Right.
- [17:37]
A
It's going to be very brand specific, I would think, at the beginning.
- [17:40]
B
Yeah. And campaign specific. And so I think there's likely going to be a need for both for a while. But I'm really excited with the pace of change and just the ability that AI is going to provide just in terms of simplicity, contextualizing even more of these signals to just provide even more value for marketers over time.
- [17:58]
A
Speaking of all that change and just a crazy pace, we are seeing in real time, just web search and web navigation change pretty quickly. And everyone wonders what's going to happen with consumers, consumer habits and shopping in particular. Are you seeing any changing changes that are impacting your business in terms of, you know, as web search really becomes more conversational or less, less traditional? I guess.
- [18:21]
B
So I think we're really anchored on how do we again provide more value to our customers, on where people are spending their time. Right. And so I think, you know, as no one really knows again, the impact of AI and, and where it's going to kind of change consumer habits the most. But we do know that, you know, or the assumption is that streaming TV will continue to accelerate. For sure. You will see continued investment in gaming and apps. You're likely to see live sports be a continued point of, you know, one of the best places to get I've heard brands talk about live sports as being one of the last best places where people are truly engaged for a long duration of time. Right. So I think for us it's really about how do we leverage again all of the signals make it easy and simple for marketers to reach people whether they're on Prime Video or the Amazon store or whether they're on Net, whether they're on whether they're on Hulu or whether they're on Disney. Right. And I think again anchoring on having more deterministic identity. So knowing as a brand that you can be more efficient with your spend because you have more clarity on who you're reaching and you're able to frequency cap and and really understand, you know, the impact of your messaging. That's going to be more important I think as you have just much more like a rapid change in consumer behavior and you're everyone's trying to kind of figure out what is going to happen from a consumer standpoint over time I guess.
- [19:51]
A
Lastly, what you know, you talked about your the pace of change and you love kind of these moments where you're building and going through these periods. What's next for you and your team? I can't believe we haven't talked about the holidays now, but what do you what's his like what should we anticipate around this the holiday season and what's coming next for you?
- [20:07]
B
Yeah, good question. I know it seems like you go into the store and every year there's something for the holiday starts three months earlier than it did the year before. Right. So it's amazing. It's right around the corner. I think broadly, as shared earlier, you'll continue to see even more investment in AI across our entire technical stack as well as more simplification, more automation and just again making it even easier for customers to quickly scale effective and efficient campaigns. I mean that's going to really be underpinning so much of what we're currently focused on, what we continue to focus on in the future. And specifically we're also excited about Amazon Unboxed, which is coming up in November. And so this is one of our biggest events of the year for those of you who've never streamed and caught it or caught it live. But we also have several innovations we're going to be revealing to customers events. So hope to see you there as well.
- [21:01]
A
Well, all right. Well that's a pretty good tease to end on. Kelly, awesome stuff great conversation and let's chat again soon.
- [21:06]
B
That sounds great. Thanks for having me. Appreciate it.
- [21:08]
A
You got it.
- [21:09]
B
Take care.
- [21:10]
A
Thanks again to my guest this week, Amazon's Kelly McLean and my partners at Elemental TV. If you like this week's episode, please take a moment to rate and leave a review. We have lots more to bring you, so please hit that subscribe button. We'll see you next time for more what's next in media. Thanks for listening.