datastackshow.com Open in urlscan Pro
104.198.107.121  Public Scan

Form analysis
0 forms found in the DOM

Text Content

SIGN UP FOR THE FUTURE OF MACHINE LEARNING LIVESTREAM!

Sign up

SIGNUP FOR OUR NEWSLETTER

Sign Up
 * Episodes
 * Events
   * Data Stack Show Meet & Greet
 * About
 * Contact

Menu
 * Episodes
 * Events
   * Data Stack Show Meet & Greet
 * About
 * Contact


EPISODE 151:


HOW TO UNLOCK THE DATA WAREHOUSE FOR MARKETING WITH CHRIS SELL OF GROWTHLOOP


AUGUST 16, 2023

This week on The Data Stack Show, Eric and Kostas chat with Chris Sell,
Co-Founder and Co-CEO of GrowthLoop. During the conversation, Chris talks about
how GrowthLoop enables marketers to build customer journeys directly on the data
warehouse. The group explores the divide between engineering teams, data teams,
and marketing, and how it is closing as more companies adopt the modern data
stack. They also touch on the challenges and opportunities for data warehouse
vendors in implementing reverse ETL, the potential for a centralized customer
data platform, the role of AI in marketing, and more.





NOTES:



Highlights from this week’s conversation include:

 * The need for reverse ETL in marketing (2:24)
 * Closing the gap between engineering, data, and marketing teams (8:37)
 * The analytics persona’s opportunity (11:53)
 * Interface layer (13:06)
 * Approach to messy warehouse data (15:57)
 * The need for a complicated infrastructure (28:43)
 * Challenges in data integration for marketers (29:26)
 * The evolution of the analytics stack (31:53)
 * Orchestration of the data warehouse (38:39)
 * The role of marketing tools (40:35)
 * Generating custom assets (46:27)
 * The shift towards making data processes easier (48:13)
 * Final thoughts and takeaways (49:23)

The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for
developers. Each week we’ll talk to data engineers, analysts, and data
scientists about their experience around building and maintaining data
infrastructure, delivering data and data products, and driving better outcomes
across their businesses with data.

RudderStack helps businesses make the most out of their customer data while
ensuring data privacy and security. To learn more about RudderStack visit
rudderstack.com.


TRANSCRIPTION:

Eric Dodds 00:03
Welcome to The Data Stack Show. Each week we explore the world of data by
talking to the people shaping its future. You’ll learn about new data technology
and trends and how data teams and processes are run at top companies. The Data
Stack Show is brought to you by RudderStack, the CDP for developers. You can
learn more at RudderStack.com. Welcome back to The Data Stack Show. Kostas,
today, we’re going to talk with Chris Sell. He’s one of the Co-Founders and
Co-CEOs and leads product at a company called GrowthLoop. And I’m so excited
about this conversation, because we talk a lot about sort of concepts, old and
new, and trends. And hot terminology in the data space. And reverse ETL is
something that’s come up on the show multiple times. But Chris runs a company
that sort of skipped the whole reverse ETL hype, and just went straight for a
use case, right? So they have a tool that enables marketers to sort of build
customer journeys and do segmentation, etc, straight on the data warehouse,
right. So it just runs right on the Data Cloud. And they refer to it as reverse
ETL. That’s a really interesting take. They also are five years old. And so a
lot older than a lot of the current reverse ETL. vendors, which is really
fascinating. So tons of stuff to talk about one of the big questions I want to
ask him, is that in terms of data engineering, I mean, you and I both know that
data engineers have been writing pipelines to get data from the warehouse to,
you know, some API downstream forever, right? That’s not a new concept. And it
makes a lot of sense to me that they’re going after a really specific use case.
And you’re actually starting to see this among, you know, the more recent
reverse ETL vendors really going after specific use cases. I want to know if
Chris thinks that’s inevitable, they sort of started there. But we’re also
saying that happened in the industry. But what do you want to ask him? Yeah, I
want

Kostas Pardalis 02:12
to have a conversation with him about rivercity l, in the context of Mark,
getting why we need to move all this data back and forth. And I think it’s, I
mean, it might feel obvious, like to people why we want to do that. But I think
that someone tries to get into the details of that and give a specific use case
where this needs to happen. It’s not that easy. And I think we have the right
person to do that. And the right use case, because marketing is one of these
parts of life, the lifecycle of a company where you really have to move the data
around by the needs is like, very clear there. Yeah. And I think that’s one of
the reasons that reverse ETL, like pretty much started from there, right? I
think most of the reverse ETL vendors that we know, like, pretty much like they
started with like a marketing use case anyway. So I think it’s going to be great
to hear about the whole lifecycle of data in this use case, why we need first of
all, like taking the data into the data warehouse would come to us like leaving
Marketo. And why then again, we need to take data out and put it back to
Marketo. And I think especially like data professionals will find these, like,
really enlightening about why we build a van like all these complex,
infrastructure and pipelines.

Eric Dodds 03:43
Awesome. I am interested in those things as well. So let’s dig in and chat with
Chris. It was good. Chris, welcome to The Data Stack Show. We have so many
things to talk about. So thanks for giving us the time.

Chris Sell 03:58
Yeah, thank you for having me, Eric. And Costas, very excited to be here.

Eric Dodds 04:03
All right, well, let’s start where we always do. So tell us about your
background, and what led you to starting the growth loop?

Chris Sell 04:10
Absolutely. So by way of a quick intro, I’m the Co Co Co Founder of the growth
loop. But long story short, what led me to this point, was actually started in
marketing over at Google. So I was a direct response marketer, dealing with the
problems of CSVs and data silos that we all know and love if you’ve ever been a
direct response marketer, and basically realizing that I had about 50 ideas, and
only five of those would ever reach any of our customers. And most of that
difficulty was because I did not have access to the day that I needed to run my
campaigns. So I was dealing with email marketing in my case, but as I started to
grow in marketing, I started to see the other channels as well. And so recently
We started a growth loop . We noticed that where a lot of the data in
organizations was being aggregated was the analytic stack. So in the Big Query,
Snowflake redshifts of the world, and they were using that for BI. And we said,
Well, why can’t marketers use that same data for marketing and create customer
journeys? So we created a customer journey platform directly on the modern data
stack to help marketers.

Eric Dodds 05:27
Very cool. Now, one thing I want to just dig into right out of the gate here, is
that what you described, a lot of people now call reverse ETL. You know, that’s
sort of like the hot terminology for, it’s kind of funny, it describes like a
general direction of data movement. So it’s not actually very descriptive
relative to, you know, all the tooling. But you intentionally said that, you
know, you built a sort of marketing automation and marketing journey platform on
top of the warehouse. So would you say that you were around you doing reverse
ETL? Before reverse ETL? Was it the reverse ETL?

Chris Sell 06:10
Yes, no, I mean, like, we kind of all work. I mean, if you were a marketer to
Eric, for your podcast here, and like, we were all sending CSVs to our end
channels to target audiences. And what I saw teams starting to do was then they
go to their analytics or Engine team and say, Hey, can you set up an airflow
pipeline or a cron job on your computer, and just automate that CSV for me?
Yeah. And you know, yeah, at the time, there was no name for what that was. Now,
it’s been given the name reverse ETL. But it is just a movement of data,
essentially, what you’re saying your marketers is, I know, you’re going to use
20 tools, I’m going to send the data to every single one of them. And yeah, it
makes sense that now with reverse ETL, there’s a more reliable infrastructure to
do that. But what we started to realize, too, was I was a business user at the
end of the day. So to me, the only thing that mattered was, Does this thing
drive revenue or an impact on my customers in a positive way? If it doesn’t, I
can’t justify it. And so as we were starting to think about growth, Lou, we
thought about it in terms of Yes, reverse ETL going from the data warehouse to
the end channels, but how do you actually get somebody to measure value and
actually produce campaigns at the same time instead of just transferring data
over?

Eric Dodds 07:31
Yep. And what? So this is really interesting, because, you know, what was it
three or four years ago that you started the growth loop? Was it almost five,
five years ago, almost five years ago? Okay. So five years ago? Man, okay, I’m
thinking back to five years ago, and thinking about, for most companies out
there thinking about sort of, you know, connecting to the data warehouse, and
directly, using data from there in some sort of downstream tool was a pretty
provocative idea. Right? And I would even say, you know, I’m not gonna make up
numbers, because who knows, but I think a lot of marketers wouldn’t even
necessarily think about that as a possibility, because of the traditional sort
of schism between, you know, engineering teams, data teams, and marketing. Yeah,
but you sort of assumed that, you know, based on your experience, can you speak
to that a little bit like, because that divide closing like, you’re on the front
lines?

Chris Sell 08:37
Yeah, I think, well, one, nobody knew what we were talking about first. So this
happened, where I think for us, it wasn’t obvious, like now I think business
applications will come to the modern data set and be built on top of it. But if
you would have asked me five years ago, I would have never made that statement.
Instead, I was at the level of trying to solve a problem for a marketer. And
actually, one of the first problems we saw was, we were working with customers,
they were leveraging Snowflake and BigQuery were the big two at the time. And
they would be using Looker, typically to do bi visualizations. And then they’d
have a data science team producing some machine learning model to predict churn.
Predicting churn was actually the first one we saw. And they were like, they did
this big presentation to their marketing teams and said, Hey, look, what we did,
looking at smarter analytics decK is that we can actually predict when our
customers are going to churn and the marketers were like, cool, like, that’s
great. Now we know that and it looks like we have several that are going to
churn. What are we going to do about it? I don’t know. Go to this BigQuery
table, we’ll do some sequencing. ESB. And so we actually started with just a
pipeline that would actually pay. We had a very simple UI where a marketer could
go and select because people are likely to turn based on that one field and
BigQuery Snowflake, and we sent that data to Marketo. And that was it. And so it
was like just solving those specific problems over time. And then marketers at
first, they didn’t understand. I mean, they still don’t really understand
reverse ETL, or the difference between a marketing cloud or what we’re doing on
a journey platform, all they care about is do they have the data they need to
run their campaigns? Faster?

Eric Dodds 10:20
Yep. How are you? This is an interesting challenge that I think will grow in
size and scope, as the data warehouse just continues to, you know, it’s like the
gravitational pull, you know, it’s gonna just continue to pull things towards
it. And even on your website, you know, you speak to both the data persona and
the marketing persona. Yeah. How do you as a, I mean, curious, you know, as a,
as kind of a business nerd and a marketer. That’s a typical line to walk. You
know, in an ideal world, there’s a great relationship there. But, you know,
practically day to day like, you know, it’s not always that way.

Chris Sell 11:05
Yeah, I think I’ll answer that question I just wanted to go back to. Is the gap
you asked me on the last one, I missed it? Is the gap closing? Are people
starting to understand it? And they are because of what you just said, when you
were saying that the data data warehouse is expanding in terms of the data sets
that it holds, as it gets more gravity in organizations and more companies are
actually adopting it. Now. They’re like, they’re starting to ask us, Why can’t
this just operate off the data stack? So that’s the first. I’ve seen that that’s
been about the last 18 months, though, and three and a half years where
everybody was just No, thanks

Eric Dodds 11:39
for educating the market for us. That’s really hard.

Chris Sell 11:44
You see these lines on my forehead? Yeah. So. But getting your question about
both personas. So if you look at it, there’s the analytics persona and the
conversations I’m having with folks, they saw the opportunity first, because
they were the ones that were pitching their organization on building an analytic
stack, and then putting Power BI Looker, or Tableau on it, because it could
solve their insights and decision making problems for their executive teams. So
they’ve already gone in internally and pitched this as, hey, this is the future
of how we get insights on our business. And so as they reach the end of that
curve, they’re like, Well, what can I do next? With all this magic, this
Goldmine, I’ve started to aggregate. And you’re starting to see the IT analytics
leader look around the business and be like, hey, I have all this data and
intelligence about the customers, what else can I solve? Now that they’re
excited to go talk to the marketing users, now the translation layer between the
marketers and their use cases in the value and its analytics stack, there’s a
translation that needs to happen. So a lot of our job really is once that
analytics leaders are excited , bringing the marketer in the room and talking
brass tacks use cases, right? And you’re playing, you know, you’re trying to
develop the relationship if they don’t have one, or make it more truthful if
they already do.

Eric Dodds 13:06
Yeah. Do you think that this is kind of a question about your philosophy? So
there’s sort of a couple of ways that let’s distill it down, because that’s
hopefully that’s even spicier. There’s sort of two ways to think about this
problem. You obfuscate technical things for the marketer. Right? And you just,
you’re essentially building an interface layer on top of the warehouse. Yeah. Or
you expose technical things to the marketer, you know, which means that you sort
of give them like deeper access, and maybe the ability to use, you know,
features that are closer to the warehouse more native to the warehouse. Do you
have a philosophy on that? Yeah.

Chris Sell 13:50
Generally, the way I think about it is most marketers, there’s a couple of
personas in marketing. There’s the data driven marketers that want to get in
there. And then there’s the 80% of other marketers that like, hey, could this up
with a bow? That’d be fantastic. Yeah. So my learning is that our interface job
is to work for that. 80%. Yeah. And to make it as easy as possible for them to
build audiences and journeys. And I don’t even want them to care what data
warehouse is running, because frankly, they don’t need to know. I don’t even
want them to understand the data schema if they don’t have to. So like audience
templates, things like that, how do you wrap this up with a bow for that
audience? But I think where, typically in business tools, what’s been the
problem, the data, that’s a SAS tool, and it’s only focused on that 80% that
means that’s your only users and the technical team ends up hating it because
there’s no interface for the technical team, right differences. When you bring
SAS to the Data Cloud. The cool part is you can service this 80% of marketers,
but then I can talk to the data teams and saying like, hey, you know, all these
audience artifacts and everything’s written back to the warehouse so you can do
analysis and you can use still use Power BI or Looker, whatever you guys want to
use. And so they still have control by virtue of you resting on this Back, they
already love the US. So like if I’m talking to other entrepreneurs and talking
to them about where I build an app, my pitch I’m building it on directly on the
stat is like you kind of get to unlock both personas, and you don’t even have to
create one of the UIs.

Eric Dodds 15:15
Fascinating. No, that is super interesting. Okay, question for you on leveraging
warehouse data. Because, you know, the idea is great, but anyone who works in a
business knows that the warehouse is generally super messy. Yeah. You know, and
there’s filling out there that that helps that, but how do you approach the
issue of like, okay, well, someone really, there’s, you know, marketers who want
to use this warehouse data, but it’s kind of messy. And so, I would guess, for
you that creates this challenge of like, well, do we try to help solve that
problem? You know, you know, the value that your tool sees? How do you approach
that?

Chris Sell 15:57
Yeah, and so there’s two, like two tactics, we typically, we’ve taken on one is,
we kind of take this approach, meet you where the data is at right now is if you
have to do a bunch of manipulation to get started and activate, that’s gonna
take time, and you want fast time to value if you’re in marketing, usually,
they’re coming to you too late in the campaign already, and they want it to go.
And so like, what we’ve done is, we have a flexible, flexible schema under the
hood. And the idea is that as long as you have a customer’s table, and some
transactions tables with a unique key on them, and events, tables, I don’t care
what columns are on them, I don’t care what they’re named, you, all you have to
tell me is how you want to join these tables together. And that’s it. And so,
one, I don’t want you to manipulate your tables if you don’t have to, right. But
two, we’re also seeing some things where you may have a common identifier, for
example, like email address, and you’re using Marketo, and Salesforce, right,
you want to join those two sources together, that’s a great one, it’s an exact
match. But we also see cases where it’s first name, last name, and they want to
join those datasets together and handle misspellings or shortened last names.
There are interesting providers coming into the space, they’re around identity
resolution directly in the warehouse. But for a large portion of our customers,
what we’ll try to do is say like, yes, that’s going to be important. But what
you can do is actually just get started with that Salesforce data today, start
building marketing audiences to expedite opportunities in your funnel. And then
you can engage one of the identity providers, and they can help you match in the
Marketo data set, which allows you to create more complex customer journeys. So
there’s always a route to get started. I think that’s actually what people miss,
is they tried to be, Hey, can I develop this perfect customer 360 data model
that all the CDP’s have promised me, but do it in my warehouse, spend two years
on it and eventually produce business value? And the truth is, your executive
teams are never going to allow you to get to yours. So you’re gonna be gonna
lose your funding before that happens. So the better approach is get the
datasets that are there and start activating.

Eric Dodds 18:06
Yep, it makes total sense. Okay, last question for me, although that’s usually a
lie. When I say that, do you think that, in many ways, it is refreshing for you
not to refer to yourself as a reverse ETL solution, but just to say, hey, we’re
building a customer journey tool for hitters that runs on the warehouse, right?
Yes, it’s going straight for the jugular, on the use case that you’re building
for with, you know, the certain datasets. But do you think that that’s
inevitable? ETL? Right. I mean, pipelines generally get commoditized over time,
you know, if it’s just, you know, sort of unidirectional data flow that is
taking data in one structure and sort of translating it to, you know, an API.
It’s just a Data Integration data pipeline problem. And so, yeah, I’m just, do
you think that the approach that you’ve taken is actually inevitable for all
these pipeline vendors?

Chris Sell 19:03
So this is going to be an interesting debate? Because the answer is I don’t
know. I think so. But I don’t know. And so we, I’ve, obviously, we’ve obviously
gone in a certain direction. And a part of that is because of our background, we
were like, We want to justify driving revenue and growth. That’s what our
business is supposed to do. It just so happens, we’re in the warehouse. That’s
how we’re gonna do it. And so we knew we wanted to go for marketing use cases.
And as we get into specific industries, yeah, that’s what a lot of our team
does. We’re talking use cases. So like in financial services, I’m talking to
them about switching people from checking into Savings Certificates of Deposit
up selling them into a mortgage. That’s the level we’re talking about. It’s
like, where are they at in the customer journey? How are you going to use your
data in the Data Cloud to orchestrate that journey? I’m not talking about hey,
do you want to get this data to Marketo? That’s not the conversation. So for
other players in the space, I think this is where the debate comes in is like I
don’t know, can there just be a strong generic reverse ETL player that doesn’t
go down the use case path and for some reason that becomes, you know, that’s
valuable as an ETL tool. I think they’re going to want to push towards value
anyways, because that’s better for the business model. But can they subsist? I
always think of the ELT tools that still have subs that are still there and
doing well, like five trans people, a fantastic partner of ours, and they’re
still growing. And I don’t. I know they’re doing ELT. But I don’t know how far
they’ve gone down the business use case path. So honestly, I’d be curious to
hear your and costus thoughts on that space? Because I think what I’m learning
is yes, they will have to get business value. But I see other examples in the
market where maybe that’s not the case. Yeah.

Eric Dodds 20:48
cost us. What do you think? I mean, you built an ETL. Company? Would you have
started going like, what do you think?

Kostas Pardalis 20:56
I think the next year is also going to be very interesting, primarily, because,
I mean, many of the things that we see like the industry out there are also
shaped by market dynamics, right? Obviously, you have data warehouses that need
the data to deliver value, like the data warehouse without data is nothing
right. And having technology and vendors like Fivetran, together with something
like Snowflake worked really well in the past. But that was like a completely
different world, like, it’s going to be very interesting to see how Snowflake,
for example, is going to react when their growth is going to start slowing down.
And they will need to figure out ways to keep growing, right? So why not
introduce some of these functionalities, for example, as part of like the core
solution, just like AWS is doing it, right. So the reason I’m saying that is
because there is a lot of fragmentation, I think like in the modern data stack.
Not necessarily with the pipelines, maybe with some other tooling out there. But
I think we will start seeing what people are calling consolidation. I’m talking
more on the product level. I think we will start seeing like there will be
reasons for Barney’s to start getting into all the stuff that their partners
were doing in the past. Right. So a

Chris Sell 22:48
quick follow up question. So like, as they hit their growth late, slow, they’re
going to be looking left and right and left is ingestion ELT, which is getting
data to my platform and reverse ETL is getting data out of my platform. So I
guess it’s a cure. I’m curious to get your perspective since you’ve built an ELT
company. Which way do you see them going first? If any?

Kostas Pardalis 23:14
That’s a good question. I will probably go to the yield T side first. And again,
I don’t think that they will just do like everything, I don’t think they will go
and be okay, let’s replicate exactly what like Fivetran is doing on top of like
BigQuery, or Snowflake, it’s going to be more like use case driven, especially
because like this combined is, I think a big part of like, their strategy is to
go after like the enterprise. And the enterprise you have to be like, are above
like the use case. Like that’s why they are saying it’s exactly what you said,
like you can go to Bank of America and say like, Okay, I’ll show you a reverse
ETL to Marketo. Right? What does this even mean? Right? Like, it’s a much more
value driven conversation, what do you need to have there? Right? So I think
neuroticism is going to be a little bit different compared to how companies like
Fivetran manage, like to grow through these big approaches where it’s more of
like, the tools out there. Get on it and figure it out, right? When I’m going to
be talking that much about like use cases. It’s about the product and the tech
itself. And I think like the ELD or ATM is going to be the first part primarily
because there’s more of us value or the data warehouse vendors they’re like data
warehousing is all about processing, right like so you need the data the gaping,
right. So especially for use cases where there’s potential a lot of data that
can come in. I think that’s like the first thing that We will see these
companies going after the other thing like the reverse ETL, where the data has
already been reduced into sub like some kind of substance and sent it back. I
think it requires much more work to figure out exactly how to connect it with
the core value of something a data warehouse delivers. So that’s at least my
opinion. We’ll see the question is also like, are they going to be building that
stuff? Or are they going to go out shopping? Yes, yes, again.

Eric Dodds 25:34
Yeah, it’s interesting to think about which way they’ll go first, to your
question, Chris. But one other interesting thing I would say in terms of them
going for the ELT side first is that the long tail of integrations for reverse
ETL is a much more fragmented problem because you’re going from like a
standardized format, say, to like a bunch of different API’s, whereas ELT is
sort of the inverse, right? Like, you are ingesting it into, like, your own
system. But the other thing is that I think that there’s I mean, if I was, you
know, planning, market expansion of Snowflake, which I’m completely unqualified
to do they have a huge opportunity to get those vendors to do that work for
them? Right. I mean, when you think about this, it is already happening,
actually. Right. So vendors are already saying, like, okay, let’s just plug
braised directly on top of Snowflake right now, how good does that work, like,
you know, for these, you know, varies vendor by vendor. But I think you’ll also
see a lot of that where the big players are just gonna be like, create us, you
know, an app, put it in the marketplace, and like you manage this. And they
offload that to the vendor, but they still have the benefit of sort of, you
know, some level of control to the marketplace. And so, I think, Cushing that
longtail fragmented, you know, API integration nightmare, off onto those
vendors, is probably wise of them as well, because they won’t spend as many
resources, trying to manage all of those integrations, without that driving, you
know, it’s like, what’s the return on how much compute that’s actually going to
drive? Yeah,

Chris Sell 27:22
That makes sense. And if they, I mean, if you do think about the data warehouse,
if that does end up being this enterprise, this single enterprise data layer,
and then all the SAS apps start coming to it? Yeah, they’re gonna want to
eventually the ELT side, whether it’s through partnerships or white labeling,
they’re going to want to make that part disappear. Because they’re going to want
the vendors of SaaS applications like Gross Lu, to just sell the value prop that
drives the computer on Snowflake. Yeah, and it just so happens, it’s on all of
this data, but you don’t want growth to go into a situation where I’m pitching
on top of BigQuery. Yet, they only have half the data there. I just want to hide
ELT and make it happen. It’s got to be there. So it’ll be interesting because I
think they’re going to try to wrap ELT in event streaming with a bow. But I
think they’re gonna. I bet it’s going to be with partners as well. It would be
my guess.

Eric Dodds 28:19
Yeah. All right, Costas. The mic is yours. Oh, mine. Wow. Let’s, I guess it’s
always been yours. Because you were, you know, you came up with the idea of the
show. But

Kostas Pardalis 28:30
yeah, true. You just put it out. It’s a lot. All right, let’s, yeah, let’s go.
Greece, like back to the basics, a little bit of like, let’s talk about, like,
the data that is needed for marketing, and why today we are talking about like
shots. Like a complicated infrastructure that is needed for marketeers to do
their job, right. And let’s assume we have, we don’t have marketeers right now
as like in our audience. But we have like these grumpy data engineers that they
get, like, all these requests about, I want this data from there, and this
pipeline doing this and that, and let’s try to help them understand why, like,
all these things are like at the end, useful and required.

Chris Sell 29:27
Yeah. Like the, the dirty little secret is probably some of them most, or maybe
most of them aren’t required. Part of this is a problem of marketers not even
knowing what to ask for because the data warehouse is a black box. So they’re
like feeling around the edges of what this thing is, which is their customer
data. And so they’re coming to the analyst team and just saying, Hey, I think
this is what I want for my business use case. Good luck. Can you write some SQL
to magically put that together, I appreciate you. And so but starting from
scratch Like, if you think about it from a marketer perspective, like, ideal
world, you don’t have to worry about any of this complicated stack, what they
would want is like you look at when you’re a smaller company, usually what
companies end up doing is they go through different stages. Now, I would
advocate for all companies. I think it’s gonna get a lot easier earlier on to
have a modern data stack right now, it’s usually large companies. But I think it
will get easier now, where people, what they’re actually doing today, though, is
marketers are starting with HubSpot, right. And they just start jamming their
customer data in there, they start doing emails, and they might use it for the
sales team for their CRM. And what starts to happen is they have their customer
data and some profile attributes, some basic things like total purchase count,
right? Or total amount spent, or last purchase date, some attribute data. And
then they start saying, like, well, actually, I want to track all the mobile
interactions we have with these customers and actually link it together. And
then Oh, actually, we’re running surveys to our customers through Qualtrics. Can
we get the survey response data because I want to run a marketing email campaign
on that. And then people are like, well, that’s not in HubSpot. Like well, how
do you get it to HubSpot? And they’re like, Well, are we asking the right
question? Maybe we should have an analytics stack. And they’re like, Yeah, I’ve
actually been thinking about using attentive push notifications to HubSpot. So I
also want you to get that data over there. So could you go get all three of
those data sets over to attentive as well as HubSpot now. So you’ve had channels
splintering, but then you’ve also had the marketer asking for additional sets of
data, and eventually that breaks. So the marketers become very unsatisfied with
an all in one solution. And then the analytics team actually says, hey, the
proper way to approach this is to have our own analytics stack. And so that
gives birth to the warehouse and an organization. And first of all they use or
is for visualization of insights. So then the marketer can ask questions, they
can see how campaigns are performing. And what we’re seeing companies. That’s
when there’s a lot of friction, though, because you’re starting to build. At
stage two, you have two centers of gravity, you have a data stack, and you have
HubSpot, let’s say you have your marketing stack and are attentive at this
point. So at that middle stage, that’s where the grumpiest of grumpy analysts
are born, including myself when I wrote SQL and people come to me and ask for
these audiences. Because this is the point where the marketer only has access to
the customer attributes that are in HubSpot, they don’t have access to any of
that mobile interaction data, they don’t have access to any of the Qualtrics
data. So what their requests start to be on the idea of Qualtrics, or the idea
of their mobile application. And they say, Hey, I really want to run a campaign
to everybody that hasn’t logged in to this section of the mobile app in the last
90 days. And you’re like, well, we then the data teams like okay, let me go
look, if that’s there, by the way, we don’t even track that marketer, let’s have
several conversations, follow ups about who you can target, then I’m gonna write
the sequel for you pull a list and why don’t you go manually load that into
attentive or HubSpot. And so that’s the stage to where there’s the biggest
friction is did you have a strong data asset, and you’re starting to build a
splintered marketing stack, and then the analytics folks are caught in between.
And so to unlock data analysts like really what when I talk to data analysts and
machine data scientists, what they want to work on is making that analytics
stack the best data layer possible for their business, for insights, as well as
activation. But then they also want to work on data science, they want to be
predicting churn or LTV or propensity models. They don’t want to be working on
pulling that latest SQL query because we’re stuck in stage two. And so
eventually, they will graduate to consider two things. They say, the marketers
come in and say, Well, we have all this problem. And I have to keep asking you
these questions. Should I buy a CDP? And the data teams like well start an RFP
process and they go start talking to all these CDP’s that say, we’ll centralize
the data for you and your stage three? And then the analytics team comes in and
is like, Are you kidding me? I spent all this time building this analytic stack.
Now you’re saying our stage three. So you have a unified customer view is the
end all another ETL adds to the CD player. So you can send your data to HubSpot
and be attentive. And so the analytics team starts to stand up to it and say
there’s new solutions being built on the warehouse, we can actually activate
this stuff through HubSpot attentive and if you change your mind tomorrow, and
you want to go with braze, we can get you over to braze because you’re going to
be able to build audiences there. And so I call that the stage three alternative
that we’re seeing a lot of analytics teams push for, because they’re kind of
tired of building up this great stack. And then it doesn’t go to us and they
have to actually just pull SQL queries or consider a CDP, which is just another
ETL to Island system. So that’s a long answer to your question, but that’s why I
think this stack exists, and is already with good intentions. They started with
HubSpot. And here we are. Yeah.

Kostas Pardalis 34:51
Wow. 100% And why am I mean, okay. Let’s, let’s, let’s think a little bit like
from the perspective of like the market here, like what would be like the ideal,
let’s say situation for the market here. My assumption and I’d love to hear your
opinion on that is that, at the end, for the market to be happy, they would love
to leave only inside HubSpot, right. Like that’s where they’re doing their work,
like they don’t want to go and use a BI tool or like data warehouse or I don’t
know, like whatever else is required for them, like go and like, create their
audiences, and then these audiences are going to be moved back and like all
these things. Are we there? today? Are we there today? Like, do you think it’s
possible for the marketeers to live only in their marketing tool of choice and
do their work there and always have like, somehow the data that they need, and
the data team can live in their own world without having to mess with whatever
is happening, like in the marketing world? Or? We are far from that? And what do
you see out there in the market? Yeah.

Chris Sell 36:09
So the ideal state for marketer, I think you’re right, it’s like they have one
place to go right. And they can run their journeys to turn wingback programs, or
whatever it is, and the channel wouldn’t matter. The reality is much more
splintered than that. So the issue, like even imagine you reverse ETL, the data
out of your data warehouse straight to HubSpot, okay, then your marketing team
started to use attentive to so then are you just going to start replicating your
database across all your marketing tools and keep those things in place, maybe,
but you’re also multiplying your data spend by 10, because you’re recreating
your customer database, and all the transactions and 10 different SaaS
platforms. So the thing preventing it is marketers should choose different
tools. Sometimes they even do it by T. So you’ll have especially in enterprise
organizations, product one and product two, one uses Marketo. The other uses
braze. That’s like good luck, right. So when you have that proliferation, what
starts to happen, and what I’m starting to see organizations go to on the
marketing side is they want their marketing teams to stop thinking as like
channel marketers, like email, or I only do Facebook ads, or I only do push
notification and attend to it. And actually think about the customer journey.
Now the issue with doing that is typically they have a splintered Mark Mar tech
stack to be able to do that. So it means to orchestrate the data across all
those channels, they have to enter like five different tools. And so what I’m
actually starting to see for those organizations, as they move more customer
journey focus instead of channel focus, is they need an easy way to use all the
customer data and go across channels. So my opinion is like essentially, I don’t
see it going all the one and destination channel for marketers anytime soon. And
it’s likely to happen because of the splintering, I think there will be this CDP
like layer that is centralized off the data warehouse where marketers will go to
orchestrate across the channels as they move more towards journey thinking then
channel thinking.

Kostas Pardalis 38:07
And how does the orchestration work? Because, okay, I totally get how the data I
think kind of works, right? Like we get all the data, we put it into the data
warehouse. We have governance, blah, blah, blah, whatever we do our magic with
the data, we have the data that we need, right? How is the orchestration
happening on the other side, though, because that sounds not very
straightforward, especially if you’re gonna be located in Japanese, right?

Chris Sell 38:39
Yeah. So from the marketers standpoint, like in our platform, and growth Glue,
it’s a journey builder tool, like a workflow tool that you typically see where
you’re saying, I’ll give you an example like Account Based Marketing, right? So
let’s say you’re targeting your Sass company, you’re targeting 10 20,000 target
customers that you deemed as these are great accounts, and I want them to know
about my business, you’re probably going to start with paid media, right on
LinkedIn ads, or Google ads. So if I can orchestrate that audience of the people
at those companies to LinkedIn ads and start advertising to those businesses,
that’s great. Let’s say I do that for seven days. I’m then going to check in my
warehouse. Well, how many signups or signup events have I received with
interest, right? So I can then say, based on that I want to trigger a nurture
campaign in Marketo. That’s a seven day campaign that sends three message
sequences to those folks based on the landing page that they went to, then let’s
say they actually sign up for a webinar. If I have that event, attendance data
in my data warehouse, I can then say, hey, that’s an intense signal. I’m going
to route that as a lead in my Salesforce. All of that can be done. Now the key
is that you have good ELT and you’re bringing those datasets in through a
Fivetran stitch. Even RudderStack is for the event stream key to bringing that
data And so you have that mobile event data, so you can orchestrate more
powerful journeys. So the orchestration can be done off the warehouse now, as
that single location to date, there hasn’t been an easy to use interface. And
that’s what we’re building for marketers.

Kostas Pardalis 40:17
So, okay, if we take the data into the data warehouse, we also create like this
orchestration layer on top of the data warehouse, right? Once you ask them the
market thing, tools, what is the value that’s like, HubSpot is delivering at the
end,

Chris Sell 40:35
creative and delivery. So I actually think creativity and delivery are
challenging problems. Now, what’s interesting to me about creative is typically,
it’s a totally different process, a different set of brand marketers doing the
guidelines, and they love to use their end tool to do it, right. So I don’t see
that going away anytime soon. They’re gonna use those tools to load their HTML
files to their QA testing processes, like that’s going to happen. What’s going
to be interesting, though, is it Gen AI is brought to the Data Cloud, which is
where a lot of these services are launched, like Google Cloud with the palm, and
now they have image Gen, and they’re going to be trying to like you can generate
creative or subject lines now, from your data warehouse for specific users. That
is very possible today, while we’re talking. So which pieces of content are
generated off the user data from your dataset versus in the platform? I think
that’s going to happen over time, I think some of that’s gonna get pulled in and
actually, towards the data stack. But I still do think most of the process lives
in that end tool. And then of course, the delivery, right? There may be tools
that end up becoming spent, you know, experts on not delivering spam and
deliverability at a certain time of day, like, these n platforms are that come
directly to data warehouse, but I think that still will remain in the channels,
right? I’ve had customers coming to us that say, Hey, I’d like to use SendGrid.
Just trigger the emails through there. All I need to do is load my templates
there. And the rest of the intelligence is going to live elsewhere. Does it go
that far? I don’t know. But it’s going to be somewhere in there.

Kostas Pardalis 42:05
Yeah, that’s interesting. And so with all these new things that are happening
with AI, how did you see marketing changing?

Chris Sell 42:18
Yeah, so it first Yeah. So some of the people on my team can tell you, I did not
think it was going to change much at all. I thought it was a hype cycle. So I
like being wrong, I guess. I do think it’s actually the real deal. And
specifically, I think, like what we’re dealing with is you’re trying to explain
data to a business user. And typically, that’s very hard, because some of us are
technical and speak SQL and database schema design. Others versus others, like
on the business side, speak in terms of personas, and what they want the
lifecycle journey to look like, I think, translation layer. And if you look at
even our product, today, we’re trying to use a UI as a transit relation layer,
we’re trying to make it easy to use the data, right? What’s even better, though,
is an LLM, saying, Actually, here’s what this data field means based on the data
in it, I’m not explaining it to a marketing persona. So I think the way I view
it as Jenny is going to attack each phase of marketing. So one is the targeting
and segmentation like how do I explain the data schema to marketers and make it
understandable so they can be self -service and easy to use? I think it’s also
going to start to be more of a chat asst of, hey, what type of journey or
sequence of messages across channels, should I be sending to turn win back users
that are most likely to win them back? I think it will still, it’ll become an
actual guide to how you do that. As a marketer, I still think there’ll be a
human in the loop there for a long time. But then, and then measuring outcomes?
How do I see what works best? Right? So which channel performs best for my
churned users on my mobile app? I’m going to be able to just ask that question.
And if I have a proper data stack, I’m gonna get the answer as a marketer. So
each of those three, I think it is going to start to infuse itself, in the way I
look at it is if Gen AI, in the LM models are all going to be in your data
cloud, and I think a lot of enterprise organizations are going to end up
training their own localized models on their own data, when that happens, you’re
going to want the predictions coming from that. And that’s going to be just
another accelerator of bringing this stuff to these business apps towards the
data cloud. Otherwise, you’re stuck with some random LLM in your SaaS tool, and
you have no idea how it was trained or where your data is gone.

Kostas Pardalis 44:36
Yeah, 100% That’s like one of the things that I also like that I find very
interesting because you see, and it’s not just like in marketing, right? Like
even like we still just like notion for example, where, okay, they bring their
own AI systems there, but it starts to feel like I don’t know like, it doesn’t
feel right Yes, like, as you said, like being able like to optimize the model
with your own data, because the data is going to become like, much more
important modes that it was before. With all these systems, I think it’s going
to be very important. And I think that’s going to be another like, like force
that is going to increase the gravity towards the data warehouses and the data
is like, whatever these data’s like is going to be right. I don’t know, like how
it’s going to look like, but there’s going to be much more gravity towards the
data. Exactly. Because of this whole AI thing.

Chris Sell 45:34
I think you’re right. And one of the unlocks, I’ve seen recently is, so some
customers out there will start, they start experimenting faster, once they can
the marketers can self serve, create audiences and measure them really quickly,
they start just throwing more shots on goal, essentially. And when that starts
to happen, the question is, how am I going to generate all this creativity, you
start moving so fast outstrip creative generation. And so I don’t think anybody
expected this. But generative AI, basically is about to change the name and
creative content generation and how you personalize it. And that’s why you’re
seeing an image in Delhi. But now I’m even seeing companies actually train the
LLM on your brand assets. And say, actually, you know, that’s not really useful
to generate a random space photo mixed with a Dali painting, although it’s cool,
that is very cool. But now I can actually generate a coke product image on a
beach, circa 1922. I don’t know, whatever. But like, now I can do that and
actually have that asset ready to go that I think is going to open up something
I didn’t see coming. Quite frankly. And I think it does make the data warehouse
more valuable, like you said,

Kostas Pardalis 46:48
Yeah, 100%. All right, Eric, the microphones bucks for your hands?

Eric Dodds 46:55
Well, the first thing I need to say is that I think there will be great
rejoicing in the marketing community, when you can get sort of LLM speed turn
around on, you know, every size of ad that you need for this campaign. I mean,
goodness gracious, it’s still so painful to do that.

Chris Sell 47:14
We were just doing a product launch working through on the marketing side. And
it was like that the hardest part was all the sizes that you need to go out to
these different channels. It’s crazy. Yeah,

Eric Dodds 47:24
yeah, that’s wild. I mean, it actually is interesting for the ad platforms that
just productize that right.

Chris Sell 47:30
Google ads just launched something like two weeks ago, basically they took their
responsive ads product. And then I think integrated palm and image under image
models and like, am I feeling lucky, you’re creative based on your goal? So
awesome.

Kostas Pardalis 47:50
So good. Okay,

Eric Dodds 47:54
last question. For the show. We’ve talked a lot about sort of reverse ETL,
pipelines, you know, layers, for the marketer, et cetera, outside of all that
stuff, and LLM, you talked about as well, what excites you in the data space,
you know, around the data cloud that you’re seeing out there?

Chris Sell 48:13
Yeah, I think what gets me most excited is it starting to reach this phase of
everybody having the architecture diagram of an ingest data warehouse with
intelligent models, and then activation out to all these channels, like I’ve
seen that everywhere. And so it’s no longer educating about what we are trying
to do together, it’s now about making it easy to do. And that’s the big shift.
So like, what’s going to end up happening is you gotta hide all this stuff.
Like, you gotta make it easier, where the technical teams can get in and get
this stuff going easily, but the marketer doesn’t seem like they don’t care. And
if they do need to care, they’re gonna keep going with the Salesforce Marketing
Cloud for in perpetuity, your go explore CDP’s. And the analytics, you don’t
want that. So I think we’re about to see a lot of innovation in the product
space. And what gets me excited as a product manager is, what’s the easy button
for doing this?

Eric Dodds 49:07
Love it? Well, Chris, thanks again, for giving us some of your time,
conversation flew by, which always means it’s a good one. And we’d love to have
you back sometime.

Chris Sell 49:17
Yeah. Thank you, Eric. Thank you, Costas. It was great joining you all today.
And hopefully, I didn’t talk your ear off.

Eric Dodds 49:23
Fast is a fascinating conversation with Chris Sal from the growth loop. And I
keep wanting to call them reverse ETL. But they don’t call themselves that.
There’s sort of a marketing activation tool, which I love, by the way, which we
talked about a ton on the show. I think one of my big takeaways from the show
that was really interesting, was we kind of had a three way conversation about
whether or not the big data cloud vendors will get into the ETL Yael space for
sort of the reverse ETL space, right? Like, are they, you know, once they sort
of start to plateau on growth, they’re gonna start looking at other market
opportunities. That was a fascinating conversation. Chris had some interesting
thoughts. You had some interesting thoughts. So I loved that one. And I think
listeners will love it as well.

Kostas Pardalis 50:23
Yeah, yeah. 100% I think, as we said, and without revealing too much, or like,
what the discussion was, the next couple of months are going to be very
interesting. And I think we are going to see movement in the market. So I think
like, our audience, like, definitely has to pay attention to that part of the
conversation. And the other part of the conversation that I think was
fascinating was when we started talking about AI. And why AI is, doesn’t look
like it’s going to be hype, just the hype, like, there is like, a lot of like
substance there. And most importantly, like, I’ve lived, like some very
interesting things, like from Greece about how AI is changing marketing, which
is very fascinating. Yes. Again, it’s one of those things where like, you can
see like, exactly, like the impact that the new technology has on something
pretty much everyone can understand. Right? Like everyone can understand like
the marketing fun. Okay, moving in. So complex, but it’s much easier to
communicate compared to, like vector databases, blah, blah, blah, like all that
stuff we usually talk about. So. It’s an amazing conversation, about the real
impact that AI has both on the market and industry and also how lots is going to
affect the data industry.

Eric Dodds 52:02
100% Also, a really good show for anyone interested in a lot of good thinking
around the interaction of data teams and marketing teams. And then also the
interfaces that need to exist for both if you’re building on the warehouse or
data cloud. It was a great episode. Definitely take a listen. Tell a friend
about it. Subscribe if you haven’t, and we will catch you on the next one. We
hope you enjoyed this episode of The Data Stack Show. Be sure to subscribe to
your favorite podcast app to get notified about new episodes every week. We’d
also love your feedback. You can email me, Eric Dodds, at
eric@datastackshow.com. That’s E-R-I-C at datastackshow.com. The show is brought
to you by RudderStack, the CDP for developers. Learn how to build a CDP on your
data warehouse at RudderStack.com.


ABOUT THE PODCAST

Each week we’ll talk to data engineers, analysts, and data scientists about
their experience around building and maintaining data infrastructure, delivering
data and data products, and driving better outcomes across their businesses with
data.

To keep up to date with our future episodes, subscribe to our podcast on Apple,
Spotify, Google, or the player of your choice.


SIGN UP FOR THE DATA STACK SHOW NEWSLETTER

Get a monthly newsletter from The Data Stack Show team with a TL;DR of the
previous month’s shows, a sneak peak at upcoming episodes, and curated links
from Eric, John, & show guests. Follow on our Substack below.

Sign Up

Powered by

© RudderStack Inc.