Latest / Signal & Noise: Bringing Clarity to Media, Data, Tech and AI / The Old MarTech Stack Is Breaking, Part I: Semantic Layers Are the New Keeper of Data and Advertising with Leighton Welch & Sarah Martinez from Tracer
Transcript
- 0:14Hey, everybody. Welcome back to Signal and
- 0:17Noise. I'm Brett House with my Co host
- 0:18Ria Longacre. Today we're taking a sharper
- 0:22look at enterprise data strategy, not as ABI or
- 0:25reporting topic, but as the foundation for AI ready
- 0:29marketing and media organizations.
- 0:31Tracer. We have two guests.
- 0:33Sarah Martinez I will introduce in a second and Leighton Welch
- 0:38sits kind of in the middle of one of the hardest enterprise
- 0:41data problems, how to unify fragmented performance data
- 0:44standards, standardized definitions and in in semantic
- 0:48layers and help teams act faster without without breaking trust.
- 0:52So it's not a black box. So to to introduce our guests,
- 0:56Leighton Welch, your Co founder, right and CEO of Tracer.
- 1:01You're the builder behind this. We love talking to the builders,
- 1:04the technical folks and getting really warm.
- 1:05We're all about that here. Yeah, Sarah Martinez, Chief
- 1:08Commercial Officer. And you, Sarah, came from IAS.
- 1:13You spent, what, four or five years at IAS?
- 1:16She's one of the favorite people in in advertising.
- 1:19Yeah. And and when when Rio introduced
- 1:21me to Sarah, I think it was a ramp up like 2 years ago, you
- 1:25know, it was, it was sort of the Jeff and Sarah and trio and he
- 1:27was like 2 coolest people in ad tech.
- 1:29And Leighton will add you to the mix 3 and so usually.
- 1:32Used to filling in just and. And Sarah, you were also
- 1:37comprised media Yahoo timing. So you've been on the media
- 1:40side, you've been on the now the vendor ad tech side, you've been
- 1:42on the publisher side, publisher side, and you're at Newhouse
- 1:46grad Syracuse. My son is currently a freshman.
- 1:51It is that you're actually trying to transfer into the new
- 1:54year Newhouse School. Love it.
- 1:56I said don't do film. You need to go into something
- 1:59that is gonna actually give you a career as creative as you are.
- 2:03And then Leighton, you being the Co founder, but you were the you
- 2:07were also at with Gary Vaynerchuk at Vayner Media along
- 2:11with Jeff, Jeff, Jeff was the chief media officer, right?
- 2:14And you were AVP of engineering? Yeah, exactly.
- 2:17My job was help bring the things that Jeff dreamed to life, which
- 2:21was a lot of crazy things, but now here we are with one of them
- 2:24is a full time company, so yeah. How do you technically it's it's
- 2:27kind of like James Brown when he was making up songs and the
- 2:30musicians are like, how do we translate that So.
- 2:33That is a good representation out of improvisation.
- 2:35What engineering is life like? Doesn't he know anything about
- 2:37music and he don't write music? And they're like, oh, well, you
- 2:40got to translate that to an E But anyways.
- 2:42But that that's that. I always respect builders
- 2:44because you guys are able mid and women are able to just take
- 2:48concepts and turn them into functioning prototypes, which is
- 2:52pretty awesome. And your background, you, you,
- 2:54you were at Harvard, you were doing, were you in computer
- 2:56science and engineering something?
- 2:58Like science and economics? Yeah, in economics.
- 3:00Which is great. We were talking before that
- 3:02you're not a true, true, You're technically a New York City guy.
- 3:06You but but you grew up. You were born in Boston, so you
- 3:08might still be a Red Sox fan. I am not actually.
- 3:11I am a Yankee fan. I grew up, you know, in the era
- 3:13the Yankees were great, right? Answer.
- 3:16Yeah, my family are are, you know, die hard Boston fans.
- 3:19So we watch sports in different range growing up.
- 3:23Yeah, you know, it's funny, ladies.
- 3:24I most of my life in New York, went to College in Boston, you
- 3:28know, in, in the belly of the beast.
- 3:29So. And everyone I've met in
- 3:32Somerville, MA on Electric Ave. at like a dive of the college
- 3:37apartment. So we have a long story back
- 3:40back to. Back 30 years so well.
- 3:42Sarah's a Jersey girl, which I just learned.
- 3:44I didn't realize that like true, I lived here, but I don't think
- 3:46I'm I can't really call myself. It's true.
- 3:48I'm at, I'm at the Jersey Shore as we speak.
- 3:50So yes, that's. Nice SO.
- 3:51So, so why don't you guys give a brief introduction to yourself?
- 3:54Just tell your audience a little bit of a you know your
- 3:56backgrounds. Sure.
- 3:57I'll jump in late and Sarah Martinez here and thank you
- 4:01Brett Maria to awesome, fabulous people in the industry.
- 4:04So right back at ya. But yeah, spent the last 20 plus
- 4:07years, as you mentioned across lots of different parts of the
- 4:11digital ecosystem. So, you know, a lot of the
- 4:13companies were focused on measurement and, and outcomes
- 4:16and insights, which everybody still is these days.
- 4:19And yeah, I mean, you know, what I kept seeing was that even, you
- 4:22know, the best tools struggled to to, you know, tell the story
- 4:26sometimes. And so teams were always trying
- 4:28to get better answers. Marketers were always asking for
- 4:30outcomes. And, you know, what drew me to
- 4:32trace her was that it's really solving the problem at the root
- 4:35cause and creating that consistent data foundation.
- 4:38And so really excited to be here with some of the smartest people
- 4:40I've ever worked with. So kudos to Layton and Jeff and
- 4:42the entire team. But yeah, it's been an awesome
- 4:45year and a half. And thank you for having me.
- 4:47Awesome. On that and I'm Layton, I'm the
- 4:50Co founder and CTO it's racer. So I oversee the product in the
- 4:54engineering org and this is the third company I've worked at
- 4:59with our other Co founder Jeff, who we were just talking about.
- 5:03And you know, I the the marketing side of this has been
- 5:07a little bit of a coincidence, but it's ended up being an
- 5:10awesome application of the things that I'm interested in,
- 5:12which have always been sort of data and process automation.
- 5:15And from day one it was very clear that there were a lot of
- 5:18marketing applications where there was a tremendous amount of
- 5:20disparate data and a lot of manual processes to automate.
- 5:23And so last couple businesses have been all about how do we
- 5:28abstract and productize that journey to go from disparate
- 5:32data to answer. And you know, we started with a
- 5:34lot of marketing use cases. And now part of what I love
- 5:36about Tracer is that we have clients where we're we're doing
- 5:39that same thing for work flows that have nothing to do with
- 5:41marketing, right. We found that the steps to do
- 5:44that are the same regardless of the question that you're trying
- 5:47to answer. And that's our mission now.
- 5:51Awesome. Well, we are really thrilled to
- 5:52have the two of you on here. And you know, we know Jeff
- 5:55couldn't make it, but you know, we've got some great, great.
- 5:58We're glad to have the two of you.
- 5:59And it's funny, I don't know if you know the story about How I
- 6:01Met Jeff. Like we actually, I think it was
- 6:04a can like four or five years ago.
- 6:06Did he ever tell you the story about commandeering the yacht?
- 6:08I mean, I think I was I was invited to a meeting like the
- 6:10last day on Thursday with someone and and then I they are
- 6:15like, oh, you can stick around. And then you suddenly you know
- 6:18how Harry from AWS showed up. Harry say I'd write it.
- 6:21He brought Jeff and a bunch of other people.
- 6:23And the next day, I know it was like 20 of us on this yacht that
- 6:25wasn't even ours that we we stayed on for a bunch of hours.
- 6:28And so yeah, that was the Jeff and have.
- 6:33You met Harry. I have met Harry.
- 6:36He's quite a Harry. He's we're going to hopefully
- 6:38get Harry in his pod soon as well.
- 6:39But but that's it. That's the introduction of story
- 6:41to Jeff and you know, you know then so great person and we love
- 6:46what you're doing to Tracer. So it's a like So besides liking
- 6:49all of you, we thought it'd be a really cool conversation today
- 6:52because data strategy, how it's shaping enterprises is super
- 6:57important right now. If you look at our smart tech,
- 6:59ad tech, data tech, like this, huge shifts going on in how
- 7:03stacks are constructed. There's been this big push
- 7:07towards composable warehouse native architecture one side and
- 7:11then like this focus on semantic layer in another, right?
- 7:13And then how are we going to take those to build better
- 7:17products? How does data governance fiddle
- 7:19in this? And if AI is really going to
- 7:20become this operating layer, right, then we need definitions,
- 7:24we need trust, and we're really going to need business context
- 7:27to really be the control plane so it can build really good
- 7:30solutions on this. And so like, because of all
- 7:33this, data management is moving up, right?
- 7:35Becoming more strategic, It's moving up the stack.
- 7:37And the semantic layer is becoming very important.
- 7:40It's coming up in so many different conversations in terms
- 7:43of its business context for AI. So knowing all this, knowing
- 7:46that we're living in a warehouse native kind of 0 copy privacy
- 7:50safe world, we wanted to think about how is that pressuring
- 7:55monolithic SAS? How is AI readiness, you know,
- 7:58how important is it? How do you, how do you prepare
- 8:00for it? Right.
- 8:01Knowing what trade tracer is doing right now, we thought this
- 8:03would be a really useful lens for understanding like what your
- 8:06company's doing, how you're helping big brands tackle these
- 8:10very important operational issues that are going on, how
- 8:13you're helping them create that shared truth that sits across
- 8:16their systems. So people in AI can really work
- 8:19on the, you know, from the same dictionary with the same shared
- 8:23logic. So that's the premise here.
- 8:25And with that, right, I think you have a question.
- 8:27Quite a theme and now we can geek out.
- 8:28It's a good one yeah. So so you and yeah, it's like we
- 8:32it's like we. Our themes are quite wide, quite
- 8:35big. So yeah, I mean, you came out
- 8:37of. These long podcasts too.
- 8:39I mean it's you can't compare 20 minutes.
- 8:41Our Joe Rogan episode where we go for like 2 and a half, 3 1/2
- 8:44hours. Again, it's just, you know, and.
- 8:46It's taking breaks in it too, like like Rogan, but yeah.
- 8:50It's a it's a skill to be able to hold the conversation for
- 8:53that long in. Two hours, yeah, there few can
- 8:54do that. Yeah, so, so sorry.
- 8:57Enough, enough banter here. But I mean, you came, you know,
- 9:00Layton, you came out of it. It's a real, this was a real,
- 9:02real world agency problem, right There was, you're dealing with a
- 9:06ton of fragmented data, fragmented teams, account teams,
- 9:10right. And you're just looking, clients
- 9:12are demanding like instant insights and like give us some
- 9:14feedback and, and there's a lot of integration issues, a lot of
- 9:17data issues. And in enterprise data
- 9:19strategies become, it's been a hot bucket, a hot button topic
- 9:23for a while. You know, I was at new Star when
- 9:26it was everybody was talking about how important this was,
- 9:28but it's really becoming, I think just much more emphasized
- 9:32now. And now that we're in this sort
- 9:33of AI agentic landscape, why do you think that is?
- 9:37Why do you think enterprise data strategy is kind of at the heart
- 9:40of of all of this? Yeah, I mean, I think AI has
- 9:43given it a reason to again, right?
- 9:45Like a lot of these initiatives have been relevant for a long
- 9:48time. The concept of like you need to
- 9:49own your own data or centralized databases like they've always
- 9:52been. I think common sense, the
- 9:54problem was just justifying the ROI for like all the cost and
- 9:58effort. I think historically it's been
- 10:00very hard to say like, oh, we did all these initiatives, how
- 10:02is it driving the bottom line? How is it impacting the
- 10:05business? And we are in a world right now
- 10:07where AI is such a hot topic that the answer to that can just
- 10:10be AI and budgets are approved and things move and things are
- 10:13happening. It's all it takes, yeah.
- 10:15Right. Like I don't need it doesn't
- 10:16even need to be hey, it's impacting the business yet and
- 10:19I'm a believer it will absolutely.
- 10:21But like we are in this unique spot where like if the reason
- 10:25that you are doing this initiative is AI, then it's
- 10:26enough. And now we are back to
- 10:28initiatives that should have been happening 10 years ago.
- 10:31What you're saying if you, if the reason you're doing an
- 10:33initiative is AI, is that that's a good thing or a bad thing?
- 10:36I think that would be. No, I think that I it's.
- 10:38Both, right? I mean, it could be, it could be
- 10:41if it's a, if it's just they have the sake of AI, right,
- 10:43which I think a lot of times it is, right?
- 10:45Yeah, I think that's why it's top of mind.
- 10:47Now. I believe that the initiatives
- 10:48that are happening to power AI or are a good thing regardless,
- 10:52right? And they're not just good
- 10:54because they are powering AI, but I think the front of mind
- 10:56and why things are getting prioritized is because AI is a
- 10:59top. Yeah, yeah, yeah, yeah.
- 11:00And Sarah, you, I mean, do you not talk about AI when you go
- 11:04into client calls? I mean, if you say that you're
- 11:06going. To just.
- 11:07We're going to show you the door now.
- 11:11No, but how do how do clients handle these kind of heady
- 11:14topics? You know, I think that it, you
- 11:17know, the biggest change that's probably happened, right, is
- 11:20that it's no longer just a reporting problem.
- 11:22The truth is it's still a reporting problem, you know, in,
- 11:26in many, many ways. And there's lots of use cases
- 11:29for, for why it's a front burner issue.
- 11:31Again, you know, even putting AI aside, when you think about
- 11:34CTOCFOCMO working more closely together, the the pressure of
- 11:38proving results, the rise of measurement, again, you know,
- 11:41all of that preparation for intelligence that we're in right
- 11:44now and that we build here at Tracer, you know, it is, is
- 11:47really what we're solving for AI here is an accelerant.
- 11:50And you know, when I'm, when I'm talking to brands, agencies,
- 11:54platforms, of course AI comes up.
- 11:56But you know, I really think we're a bit back to basics too,
- 11:59right? And so the front burner issue of
- 12:01if a being a front burner issue again is a good thing because we
- 12:03need to get back to base, back to basics.
- 12:05We need to get back to cleaning up the basement so that you can
- 12:08be innovative as a brand, as a marketer, but also just to get
- 12:12faster insights because the pressure is real right now
- 12:14across that those 3 organizations.
- 12:16Do you get asked about AI and you have to like back the
- 12:18clients into that Sarah where he's OK?
- 12:20Hey, listen, like to do this, you need you're going to need
- 12:23these basics, you going to need to get your dad in good shape.
- 12:25You're going to build need to create those, you know these
- 12:27different layers like education. Yeah, there's an education
- 12:31component, 100%. I do think that more CMO orgs,
- 12:34marketing leaders, you know, the way that I'm seeing brand
- 12:37structuring their teams now, that piece of education is
- 12:41becoming more and more important.
- 12:42And I think in the future, those marketing orgs are going to
- 12:45truly have to understand data infrastructure to be able to be
- 12:47leading in the AI space. But we start every conversation
- 12:51with an education of the landscape, 100%.
- 12:55Or, or do you start of a conversation with like this is
- 12:57your fundamental problem now we're going to get you about the
- 12:59landscape. And some exactly, it's a mix of
- 13:02both. You've got a lot of you have,
- 13:03you have a lot of customers out there who are hand raisers and
- 13:06say I'm owning that we have the challenge right.
- 13:07And then you've got others who need to understand that they
- 13:10have it and why. And so it's a pivot for us as a
- 13:13as a company to talk to customers in that way.
- 13:15But you know, it's constant education across all.
- 13:18Kind of mapping the two in a sense you're connecting the two,
- 13:20you got a problem. AI is not just hand waving, it
- 13:23actually can really accelerate, let's say decisioning or
- 13:25prediction or insight. And this is how yeah 23.
- 13:30It's collaborative. Like, I think that's the biggest
- 13:32thing, right, is these large enterprise organizations can
- 13:34sometimes be, you know, a bit siloed as as we know and, and
- 13:38including, you know, many organizations we've probably all
- 13:39worked in, which is very normal. But you know, we call ourselves
- 13:42a collaborative analytics platform for a reason.
- 13:45Siloed data, siloed team, siloed decision making is all something
- 13:48we have to solve for on the day-to-day when it was a
- 13:51reporting problem and still is. But it's also now becoming a
- 13:54bigger problem knowing, knowing AI is is here.
- 13:56Yeah, it makes sense so late in looking at like some of the
- 13:59themes I called out earlier. And just to refresh here,
- 14:03semantic layers, warehouse native architecture, 0 copy
- 14:06activation, governance versus infrastructure, like what jumps
- 14:11out to you out of those or other things you're hearing is like
- 14:13the most urgent now when you talk to brands that they're
- 14:16telling you things you need to solve, whether you're telling
- 14:18them maybe to Sarah's point, you're recommending they they
- 14:20tackle what jumps out. Yeah, probably some, I, I think
- 14:24semantic layer would be the one that jumps out most to me.
- 14:26I mean, I'm, I'm biased because that's a lot of like where we
- 14:28focus. But I do think that there are a
- 14:30lot of great tools already in the the warehouse space.
- 14:33Like companies like Databricks and Snowflake have taken off for
- 14:36a good reason that they provide a lot of those other services,
- 14:38right? They do provide tools for
- 14:40governments for for copy for a lot of things that you just
- 14:43talked about. And what's largely missing is
- 14:46the translation of that data to business concepts that become
- 14:49accessible to downstream consumers, which you know, is
- 14:51often AI agents, so. I think you just define
- 14:55semantics or not semantics, but this semantic layer, right,
- 14:59right, which I think in syphilis terms, and I'm going to say this
- 15:01non-technical and then you correct me, Leighton, is that
- 15:03it? It is the translate.
- 15:04It's sort of the do's and don'ts and guard rails of how they
- 15:07should interpret data Beneath it is a middleware.
- 15:10Is that is that an accurate? Non, I think that's totally
- 15:13accurate. Yeah.
- 15:15And so, you know, that's a lot of what's missing.
- 15:18And I think, you know, Tracer provides a little bit of a
- 15:20unique approach to that semantic layer where today a lot of
- 15:24organizations are doing that really one or two places.
- 15:28They're doing it downstream in ABI tool, right?
- 15:31And I think there are some great tools that are, you know, new
- 15:34wave BI companies that are starting to tackle that.
- 15:36But part of the problems that if the semantic layer doesn't live
- 15:38in your source of truth, then you're recreating that knowledge
- 15:42in multiple places, which defeats the purpose.
- 15:45And then I think like having that data in your warehouse is
- 15:48important to to send it to other places, but the interface to
- 15:51interact needs to be something accessible.
- 15:53And So what I mean by that is like if your semantic layer is
- 15:57writing code in the warehouse, right, then the people who have
- 16:00the business context can't directly contribute to it
- 16:03because they're not engineers. And so bridging that gap is a
- 16:06lot of where we live of like how do we put AUI on top of this so
- 16:10that the people who understand the context of this data can
- 16:13contribute to the semantic layer rather than just be consumers
- 16:17of. It and conversational with
- 16:19conversational type inputs like with the real world.
- 16:21I think with conversation through AI, like we provide
- 16:24tools to do that. Also just through AUI, that's,
- 16:27you know, that's drag and drop, that's accessible to people who
- 16:30aren't traditional data people or engineers.
- 16:34Yeah. So I think good, I think
- 16:37somebody's phone is is beeping, but that's why.
- 16:39Yeah. So so I read a report that I
- 16:41thought would actually be interesting.
- 16:43And if you guys haven't checked out, it'd be interesting to talk
- 16:45about and sort of get your perspective on both Sarah and
- 16:48Layton. Scott Brinker, who I'm sure a
- 16:50lot of you guys follow, just did a report with data bricks that
- 16:53one of my colleagues, one of the Co founders of high signals sent
- 16:57over to me. And it's called the new Martic
- 16:58stack for the AI age was written by Brinker, but sponsored by
- 17:02data bricks. And he he did a good report.
- 17:05Yeah, it's a really good report. And it's it's like 36 pages, but
- 17:07I read every line. It was like, wow, this is this
- 17:09is kind of what a lot of we'll be talking.
- 17:11About you just use AI to summarize it.
- 17:14And I wrote it. That's what I did.
- 17:15Publishing it Monday, but basically it describes 3 AI, 3
- 17:19martech ages, which many of us have lived through, right?
- 17:22And we're in the third right, the first just sort of the 90s
- 17:25to the 2000 tens think like monolithic SAS, large standalone
- 17:29products, Salesforce, Adobe, right.
- 17:32And then often times those products get combined with other
- 17:34products to make sort of multi product suite.
- 17:36So that's that's a world that we still sort of live in.
- 17:39It's still there, right? The second age is sort of 2000
- 17:42tens to 2020. Other, the last one peaked
- 17:46around 2018 I would think, right?
- 17:47Yeah. So there's some overlap between
- 17:49these, but it's the other that makes sense, right?
- 17:51And it's sort of increasingly open systems, big talk around
- 17:54interoperability and and API integration thinks CDP's even
- 17:58even the birth of composable which we've seen my brother
- 18:01works for a company in Boston called Elastic Pet, which is a
- 18:04composable commerce company. So that topic composable CDP's
- 18:08like high touch is huge and we're still sort of in that era,
- 18:11but the now and kind of going forward, what he calls the 30
- 18:15era it AI sort of proven to be the catalyst, right?
- 18:18And and his whole concept is that Mark tech no longer sits on
- 18:21top of the data. It is the data, right?
- 18:25And and that seems to me like because it's driven by a Gentek
- 18:29AI. Can you talk a little bit to
- 18:31that? And then you probably think a
- 18:32lot about these high saluting concepts late in terms of in
- 18:37terms of agentic and how it how it lives on top of the data and
- 18:40it becomes almost it's sort of like destacking the stack,
- 18:44right? Yeah, great.
- 18:45I think that the idea here is you don't want all of these
- 18:48things in different silos. Like traditionally even, you
- 18:50know, you have a martech tool, you have an CR and you have an
- 18:55ERP, right? Like in those things, answer the
- 18:57business questions end to end for the departments that are
- 19:00using them, But then you want to go trust department, you're at
- 19:03the same underlying problem with building a stack that connects
- 19:06those things. So like that problem was still
- 19:09there. I the idea of having one
- 19:11technology that could like answer and do everything for
- 19:14every department was never really going to happen to begin
- 19:17with. And I think the, the agentic
- 19:19piece of this is, is a couple. I think one, it's easier now to
- 19:23draw insights on disparate sources than it ever was.
- 19:26And two, it's easier to build narrowed applications than it
- 19:30was previously. And so maybe you're automating
- 19:33very specific work flows or tools are popping up.
- 19:35They can do very specific things.
- 19:38And as long as you have this central warehouse that can, that
- 19:41can read and write the data that is being produced by that
- 19:44application, then it's not inherently a bad thing, right?
- 19:48And, and that's really the infrastructure that needs to
- 19:51happen where it's like if I have tools that are doing very
- 19:53specific things, the context and data coming out of that needs to
- 19:57be sourced to some centralized location so that other tools can
- 20:00contribute and learn from that knowledge base.
- 20:02And so like, that's where I see things going right now.
- 20:05Yeah. And that eliminates like data,
- 20:07data duplication, data silos, right?
- 20:10You know, there was a lot of client problems we solved at new
- 20:12start where there would be GM at.
- 20:14I've said this like 6 times but I think I always think it's just
- 20:17a fascinating point is they had 60 different data warehouses.
- 20:20Yeah, I worked from all my clients where it was similar,
- 20:23Brett, where there's just like. 33 or 4.
- 20:25Versions of the same data, different living places.
- 20:28So I mean, I guess looking at that like like what would be
- 20:31like, you know, sorry, I guess what are you hearing from
- 20:33clients in terms of the reluctance to copy data to
- 20:37replicate it to like, well, like what are trends you're seeing?
- 20:39And I guess like how does like Tracer playing in the in that
- 20:43space increasingly? Yeah, I, I think it comes a
- 20:46little bit back to education. So when we enter the market with
- 20:49a customer and we'll, we'll go the marketer side, the brand
- 20:52side, sometimes we're talking to the CMO org specifically, right,
- 20:55because it's so fragmented and they're dealing.
- 20:57Typical buyer is a is a CMO or someone to marketing.
- 20:59Yeah, absolutely. I mean it, it's definitely in
- 21:01that org. However, you know, I will say, I
- 21:03mean, you know, we've got a, an agency that we work with and you
- 21:08know, the CFO, you know, will specifically say in the last 35
- 21:11years, I've never been able to see what I'm able to see, which
- 21:13is, you know, all of the data in one place to be able to actually
- 21:16have their own Billings off of actual media data.
- 21:18Like there's so much more to it from a customer perspective, but
- 21:21when we get in the door, it's usually to the fragmentation of
- 21:25the CMO org because they're living it and breathing it and
- 21:28they've got to answer to the CFO, right?
- 21:30In many ways, once we're in, we see many different business
- 21:33operational use cases that that, that, that, that that come
- 21:37together in terms of helping. So but going back to the
- 21:40question just in regards, I think it just keeps coming back
- 21:42to education. There's many people that don't
- 21:44realize they have multiple warehouses, right?
- 21:46And all they need is the outcome.
- 21:48And people in marketing probably don't know, right?
- 21:50Yeah, they don't realize it as much.
- 21:51And then then they pass us to, which is great.
- 21:53They'll pass us to the analytics team and the analytics team and
- 21:56the engineering team who start to explain to us, you know, what
- 21:59they're working on and and you know how they're trying to solve
- 22:02for the challenges of speed and flexibility and whatever that
- 22:04might be. But usually when we end sure
- 22:07that there's not necessarily an understanding of some of the
- 22:10duplication that's happening. And I do think that's got to
- 22:13change. I think that organizations got
- 22:16to be well, it might be we have. ACVP here.
- 22:18We've actually another CVP here. Then we have a Gmail platform
- 22:22here, here. Then we have like a reporting
- 22:24data Mart here. Might be the same data copied
- 22:25over multiple times, right? Yeah.
- 22:28And you're saying the team's often times the CMO teams,
- 22:31they're often times without the analytics teams don't even know
- 22:35all of the nuances of the stack? No, of course I mean, not the
- 22:40infrastructure stacks so much sometimes.
- 22:41Absolutely not. Like they know they have a
- 22:43challenge, they get it, but they don't understand the nuances
- 22:46behind it. You know, they don't truly
- 22:48understand why it takes so long to get to XY and Z output that
- 22:51they might need and so. Quickly, do you have to pull in
- 22:54the technical, analytical folks to to help you diagnose the
- 22:58problem? Because if they're not aware of
- 23:00what the problem is. Or you could pull in a
- 23:03consultancy to help. Yeah, yeah, yeah, that is very
- 23:06true. And no, I mean, we, they get
- 23:08pulled in pretty quickly, right. So it's almost like CMO org
- 23:10validates the fact they have the challenge, then they're like go
- 23:13talk to our analytics team, then they're all learning together.
- 23:16You know, we're bringing knowledge to the entire
- 23:18organization in many ways. And then from there, you know,
- 23:21typically right now we're hearing a lot about, you know,
- 23:24API ingestion and keeping up with, you know, maintenance of
- 23:26of connectors, right? We're hearing a lot about we've
- 23:29got everything in our warehouse, we're good.
- 23:30But then in the warehouse, it's a mess.
- 23:32And So what does that mean, right.
- 23:33And so I think everyone's kind of like loud and proud that
- 23:36they've got everything in the warehouse sometimes.
- 23:37But then they all of a sudden, you know, sit there and go, OK,
- 23:40but wait, it's my closet. And it's really messy and it's
- 23:42not organized by size or season or color.
- 23:44You know, what do I do about that?
- 23:46And that's where we can come in to truly educate what that
- 23:48process looks like to from preparation to output, if that
- 23:51makes sense. One sense I get like maybe late
- 23:54in this night before you is like I like, I mean like the pendulum
- 23:58swings back and forth, right? Like I think for a long time
- 24:00marketing data was run by IT. Then I think with the rise of
- 24:03SAS like this started becoming democratized where marketers
- 24:07could like just use a credit card and get a new platform and
- 24:10upload their data. Next thing you had this sprawl
- 24:13like the SAS sprawl with data being sort of leads from places
- 24:15that we're just talking about. I kind of sensed it like the
- 24:18center of gravity or the pendulum swinging more back
- 24:21towards IT these days as the like the cloud data warehouse
- 24:24becomes more important and becomes kind of recognized as
- 24:28let's say a best practice for architecture.
- 24:30Not the only pattern, but like a best practice for most
- 24:32enterprise organizations, right? Is IT typically like, do you
- 24:37agree with that number one? Like what are you typically
- 24:39running into when, when you work with your clients?
- 24:42Are these IT decisions? Are they IT plus marketing
- 24:44decisions or marketing's consultant?
- 24:45Are they informed? How does it typically work?
- 24:48No, I agree. I think that has been the the
- 24:50pendulum swing. I think it's definitely back to
- 24:52a centralized kind of IT owned. This is where our data lives.
- 24:56And I think for all of our enterprise clients, there's some
- 24:58version of that. I think it's a good thing in
- 25:01that like at the end of the day, marketing data in a silo isn't
- 25:04super useful. Like it's all about driving
- 25:06business outcomes anyway. And so those things need to be
- 25:09buried. I think where we need to be
- 25:11careful now as it swings is that you're not creating a humongous
- 25:15bottleneck, right, where every decision has to get approved
- 25:19through that very arrow funnel. And every time marketing wants
- 25:22to do something, every time finance wants to do something,
- 25:26every time a product person wants to do something, it's
- 25:29getting funneled into. A ticket and JIRA, then you're
- 25:31back in the key, right? So I think that the gap there is
- 25:35going to be like, yes, there should be one source of truth
- 25:38for data, but in terms of access and consumers, it can't be a, a,
- 25:42a narrow funnel like a bunch of things need to be able to sit on
- 25:46top of that warehouse, read and write to that in a, in a way
- 25:49that doesn't slow. Everything down well and in
- 25:52teams, functional teams have to have equal access, right and
- 25:56equal education and internally to like how what, what kind of
- 26:00insights, what can the insights can the CFO derive from this
- 26:03through an agent or otherwise the CMO otherwise, right.
- 26:06Often times that's kind of a fragment at play or one
- 26:08department's using one part of the stack and doesn't really
- 26:11have access to that same insight information.
- 26:13How do you guys handle that organizational challenge?
- 26:16The friction between teams, Some teams are in the end, some are,
- 26:19are not right politics. I mean, how do you?
- 26:21That's got to be the hardest part of the job, Sarah.
- 26:24You know, I think I mean there's, there's hard things,
- 26:28but at what, you know, what I would say is that a lot of the
- 26:31times when we're walking into some of the larger marketers,
- 26:33it's it, it can sometimes be use case driven, right.
- 26:35So you know, when it's use case driven, then it's sitting within
- 26:39the marketing org that's an executive priority saying we
- 26:42need to, we need faster insights from our measurement models,
- 26:44right, which is the biggest conversation right now.
- 26:47And so with that, when it's an executive priority, when it's
- 26:49sitting within the marketing org that has the PNL to do it and
- 26:53it's going to have impact for the CFO as well as all the brand
- 26:56marketers and all the different regions that they're in.
- 26:59You know, it's then not that hard of a conversation, right?
- 27:01It's OK, we're aligned on this use case.
- 27:04Knowing how important this is for you, let's focus on that.
- 27:07Then once we're in the door, and we've now taken a process from
- 27:11six months to two weeks for them and they're now getting
- 27:13measurement model, you know, outputs back on on the regular.
- 27:17Then when they start talking about things like how can I
- 27:20bring trade data and how can I bring sales data and how can I
- 27:23bring in DMA level data across everything so that I can tie it
- 27:26back to, you know, the franchisees that are asking me
- 27:29for things, right. So that unlock really starts
- 27:31once they start to see the power of collaboration across all of
- 27:35the teams. And it takes time.
- 27:37It's change management at the end of the day, but you know,
- 27:40once we are, we're in the door with, with the use case in
- 27:43particular, it's, it's pretty obvious to everyone how powerful
- 27:45not just the tool is, but the opportunity to, to, to drive
- 27:48insights so fast. And are you guys providing like
- 27:50the the attribution models or is that in partnership with
- 27:53somebody else and you guys are managing more of the data
- 27:55unification, data normalization decision?
- 27:58The second. Yeah, yeah.
- 28:00Got it. Yeah.
- 28:01We're not a recommendation engine.
- 28:03We are agnostic to every partner that we work with so.
- 28:05You could feed into a recommendation engine or
- 28:07decision. Yeah.
- 28:11And having been on the on the MTA and MMM side with market
- 28:13share and Nustar the the data unification, that data problem
- 28:18was the biggest problem and it's what delayed our ability to
- 28:21deliver any insight at all. So you had the sort of time to
- 28:24value was just. Exactly.
- 28:26In the in the in the tank because it took six months to
- 28:29get the data in order in order for us to run the model to
- 28:32actually give. Wasn't even relevant by the time
- 28:34you got the insight. Yeah, especially if it's like
- 28:36batch data, it's, it's getting completely outdated, Yeah.
- 28:39Exactly. And that's exactly right.
- 28:40Hear it from the measurement partners and from brands and
- 28:43agencies. But time to value, speed to
- 28:44value when you're spending millions of dollars in a in a
- 28:46clean room in ACDP, whatever it might be that they want insights
- 28:49right away. And so it takes, it takes so
- 28:51much time and that is something that we're is very core to our,
- 28:55our use cases. Because you got to translate
- 28:56those insights into recommendations and they
- 28:58translate those into actions. I mean, even that takes a bit.
- 29:01So I can see why getting those quicker, more timely would be
- 29:03super important. I guess I'm looking at Tracer
- 29:05though, like now that we're, we pivoted there, your model
- 29:10knowing that it, you maybe can walk our viewers through this.
- 29:14So like ingestion, preparation, intelligence.
- 29:17Like how would you describe like what Tracer does and where does
- 29:21this sit relative to other tools like data warehouses, BI tools,
- 29:24EI apps? Maybe explain that a little bit.
- 29:27Yeah. So I mean, we touched on that
- 29:29composable word before and that's a big piece of our thesis
- 29:32because one, we don't want to try to fix things that aren't
- 29:36broken, right? Everybody is coming to us with a
- 29:39different problem in that data workflow.
- 29:41And so you know, you mentioned ingestion, preparation,
- 29:44intelligence, Those are the three phases that we trying to
- 29:47produce through our product. And some clients are using all
- 29:50three. Some of them we are sitting on
- 29:52top of their warehouse, providing that context layer,
- 29:55that preparation and then sending that data to their BI
- 29:59tool, their AI agents, their warehouse, etcetera.
- 30:03But quickly, I think the way I think about ingestion is like
- 30:06this is our like skip to their base for clients who haven't
- 30:10already built it out. And so if you're, you know,
- 30:13we'll use the marketing use case, if you're running, you
- 30:17know, on 20 different EDA platforms, either have already
- 30:21built out this ETL pipeline of like I use a data partner A5
- 30:25Tran or an air byte or whoever it is.
- 30:27I have a transformation layer, you know, using maybe DBTI
- 30:30orchestrate with Airflow, like monitor with data doc, etcetera,
- 30:33etcetera, right? Like you've basically pieced all
- 30:35that together for them so that they can just use Tracer instead
- 30:39of put it together. And if they've already built it,
- 30:41great, then we sit on their warehouse and we work that way.
- 30:45The preparation layer is, is the semantic layer, right?
- 30:49It's how we go from all of those data tables to things that
- 30:52provide business value. So how do we organize every ad
- 30:56that you've run into the, the market, the product catalog, the
- 31:00the creative category, the audience segment, right, and
- 31:03normalize all of those things. How do we define relationships
- 31:06between data sources? What are the KPIs that matter?
- 31:09And then all of that, it's into this intelligence piece, which
- 31:12is hey, we're providing the semantic layers into the models
- 31:15to produce feeds back to the warehouse to answer business
- 31:19questions in Tracer, to build reports or decks or whatever it
- 31:21is. And so that's kind of the the
- 31:24phases of the product attempt. And so that last step when once
- 31:28it's pushing to, it'll push to an external BI tool tool and
- 31:31external analytics tool for either dashboarding and and MTA
- 31:35tool for so that, so that's what you're doing is you're pushing
- 31:38that to a tool to actually get an output.
- 31:42Yeah, I mean, some clients are just answering questions
- 31:43directly in Tracer and it never has to leave, right.
- 31:46And and others it, you know, if we're working with an MMM
- 31:49company, then like the end results are the contextualized
- 31:52data sets that are feeding their models and, and we work both
- 31:54ways. But I think the nice thing about
- 31:57the way that we're set up now is that insights don't have to live
- 32:00in one place, right? Like you can answer questions
- 32:03using tracers, AI and your own. And as long as it's being fed
- 32:08the same data sets and has access to the same semantic
- 32:11layer information, it should be able to produce insights in a
- 32:14relatively consistent way. Right.
- 32:16And that's the goal versus I think, like, historically,
- 32:19there's been a push to be like 100% of our dashboards need to
- 32:22be in Tableau. Yeah.
- 32:24And like, maybe that makes sense for some of the things and not
- 32:27all of the things. And I think now like that,
- 32:28bifurcation is actually possible and people are far more willing
- 32:32to do it. Makes sense Earlier on we
- 32:34touched on semantic layer and it's just a term that's coming
- 32:37up a lot these days because she says a hot term, right.
- 32:42Looking at that, Sarah, like, like, I guess thinking that like
- 32:47the semantic layers is really the context layer, right?
- 32:50It's going to you're going to build that you're going to need
- 32:53in order to build AI applications.
- 32:55How does Tracer like how would you describe Tracer's role in
- 32:59terms of that? And like is Tracer's tracer
- 33:02effectively become the control plane for marketing and media
- 33:06definitions and then you would build AI models and applications
- 33:08on top of. Is that how you would describe
- 33:10it? Yeah, I, I definitely think so.
- 33:12I think I almost look at it as, you know, you know, so every,
- 33:16every marketer's got different definitions for, you know, even
- 33:19within their own teams, they have different definitions of
- 33:21for certain things and the way reports are coming in.
- 33:24But I think overall it's really, yes, we're creating that data
- 33:27dictionary with them, right. We absolutely make
- 33:30recommendations on it. You have some marketers who are,
- 33:34you know, need a lot of help on the taxonomy and data definition
- 33:37front. You need others that are halfway
- 33:38there. You have others that have got it
- 33:39completely cleaned, right. But if it was ever perfect, then
- 33:42we wouldn't really have this challenge as much as we do
- 33:44today. But yeah, absolutely.
- 33:46I mean, I would, I would 100% look at us as that, as that
- 33:49Control Center for them. But it's really about at the end
- 33:51of the day that at that point, when we're making the data
- 33:54dictionary, it's about people and process, right?
- 33:56It's about really getting everybody together to truly
- 33:59understand what's the challenge we're solving for?
- 34:01Why is the challenge happening, Which usually has to do with
- 34:03Excel spreadsheets and you know, you know, people putting in
- 34:06things differently. But having that baseline
- 34:08playbook with the companies that they understand why that shared
- 34:11language and semantic layer and business context layer is so
- 34:14important is a really big part of, you know, the conversation
- 34:17when we start our our engagements.
- 34:19And then really then getting into just data access in
- 34:21general. Like you'd be surprised.
- 34:22There's a lot of a lot of marketers who who don't
- 34:25necessarily always have access to their data, right?
- 34:27It's theirs, they don't have access.
- 34:28You know, sometimes agencies do, sometimes it's an in house team,
- 34:31sometimes it's a team over in IT.
- 34:33And so we're really starting at the foundational level to bring
- 34:36it all together, to be really collaborative to get to that
- 34:38semantic layer. So there's some pre work that
- 34:40happens, but we come in with a perspective of viewpoint, our
- 34:44experience. We're working with many
- 34:45marketers and agencies to help them on their way in that.
- 34:48A lot of these requests is it is that they're looking to cut
- 34:51costs, is that they're improved. Looking to improve governance is
- 34:54like we just need to do something AI we need and we
- 34:57realized we were missing these definitions of the semantic
- 34:59layer. What's driving a lot of the like
- 35:01the conversations you're having today?
- 35:03I would say, I mean, you know, as mentioned before, I, I, I
- 35:05definitely think, you know, you know, measurement, you know, it
- 35:10is definitely the biggest use case of everybody realizing they
- 35:13can't get to their outcomes fast enough, right?
- 35:15They can't prove out the value of their media fast enough.
- 35:19And so that's a good like a very important use case.
- 35:21AI obviously is a Halo effect across all of our conversations.
- 35:25But I I also think sometimes it's just like they're realizing
- 35:28the value of what they're sitting on then and they can't
- 35:31believe that they can't extract it sometimes.
- 35:33We were working with an agency recently, he said, you know,
- 35:35we're the largest buyer of X, you know, programmatic platform.
- 35:39And I want to be able to, to, to share with everybody all these
- 35:42incredible benchmarks and insights that we have being the
- 35:45largest, you know, seat on that platform.
- 35:47And so how can I do that, right? And, and how can we start to do
- 35:50that from a bottoms up and top down lens?
- 35:53That's a data semantic issue, right?
- 35:55It's a, it's a data foundation issue.
- 35:57And so there's lots of different ways.
- 35:59I think it's a very strategic conversation, but typically
- 36:01where it starts is about outcomes is about, you know, how
- 36:05do I reduce, you know, time to value and how do I make sure
- 36:08that I can prove out that I'm buying the right media at the
- 36:11right time and, and have a message in front of the right
- 36:13person. Well, and it sounds like it
- 36:14sounds like education is a big key to to client stickiness and
- 36:18client lock in, right? Like if you're if you're
- 36:20educating the CMO and maybe this is my questions leading towards,
- 36:23you know, CM OS are are have to become more data flowing and
- 36:26more technical by definition. Are you seeing that a lot of
- 36:29them are saying this is this is my career future?
- 36:32You, you know, please educate me.
- 36:33Please help me understand what a semantic layer is, how it fits
- 36:37into display because knowledge is power in these situations.
- 36:40100% and I definitely think the curiosity has increased
- 36:44tremendously. I mean, as I mentioned, I was
- 36:45only only here a year and a half and even in just this year and a
- 36:48half, the questions, the curiosity, the need to want to
- 36:51know the, you know, all of that has 100% seen a a positive
- 36:55shift. I also think they're hiring
- 36:56people that have that understand it right.
- 36:58And so, you know, you know, and, and, and you know, what I'm also
- 37:01learning is they're putting people in roles within
- 37:04organizations that aren't necessarily so fluent
- 37:06necessarily maybe on the marketing front, but they're
- 37:08super fluent in the technology front, right?
- 37:10And so, you know, I think just being able to optimize the
- 37:13organizations you have and structure them in a way to be
- 37:15able to lead and all learn together is something that I've
- 37:18seen as a big change to that's been happening.