00:00:00
Hello and welcome. You're listening to
00:00:03
Marketing Matters on the Wharton Podcast
00:00:05
Network, which is our weekly podcast
00:00:08
where we analyze the latest in
00:00:09
advertising, marketing, customer
00:00:11
behavior, new product launches,
00:00:13
retailing, branding, anything marketing.
00:00:16
I'm Barbara Khan, the Patty and JH Baker
00:00:19
Professor of Marketing, and I'm joined
00:00:20
by my co-host, America Reed, the Whitney
00:00:23
M. Young Jr. Professor of Marketing and
00:00:26
the Brand identity theorist. Hello
00:00:29
Americus.
00:00:30
>> Hello Barbara. So last week we sat in on
00:00:33
a presentation in our department uh
00:00:36
professor by the name of Kevin Lee,
00:00:38
University of Michigan marketing prof
00:00:39
talking about AI and how AI can be used
00:00:43
agentically
00:00:45
to help design your emails and outreach
00:00:50
and protocols and different collaterals
00:00:52
for your marketing efforts. and they
00:00:54
actually introduced this idea of having
00:00:57
AI sort of help you figure out in AB
00:01:00
testing what's working, what's you know
00:01:03
what's not working and so on. So really
00:01:06
>> really cool result. They actually
00:01:07
reported a 33% lift. So now I'm
00:01:10
wondering I I want to go deeper on this
00:01:12
idea of of agentic AI and like the the
00:01:16
marriage Barbara between these new tools
00:01:19
and us as marketers trying to measure
00:01:22
stuff trying to elevate and amplify. Is
00:01:25
there anything you've got for me that
00:01:27
can help shed some light on this today?
00:01:29
>> Why don't we get someone who really
00:01:30
knows what he's talking about? Uh, and
00:01:33
I'm really happy to have with us in the
00:01:35
studio today Chris O'Neal who's the CEO
00:01:39
of Growth Loop. And Growth Loop is
00:01:41
trying to redefine that future of
00:01:44
marketing using, as you're suggesting,
00:01:46
AI and data. Chris, welcome to our show.
00:01:50
>> Hi, Chris.
00:01:50
>> It's great to be here with you.
00:01:52
>> So, Chris, before you tell us about the
00:01:54
world of AI and marketing and how to
00:01:56
improve everything, can you tell us your
00:01:59
background, how you got to where you
00:02:00
are?
00:02:01
Yeah, absolutely. I grew up in a very
00:02:03
small town of 7,000 people in the middle
00:02:06
of nowhere up in Canada and I found my
00:02:09
way to the valley for two months uh
00:02:12
before graduate school and that's turned
00:02:14
into 25 years. I've been very fortunate
00:02:16
to be part of some incredible companies
00:02:18
like Google, some amazing teams and
00:02:20
brands. Um and really uh the first wave
00:02:23
of that was really trying to figure out
00:02:25
how you actually use this thing called
00:02:27
digital marketing to help um build
00:02:30
personal connections or just connections
00:02:32
with with consumers and businesses. So
00:02:35
that was the first wave and now we're uh
00:02:37
here an exciting new wave with applying
00:02:39
AI in uh in an even more transformative
00:02:42
way. So I'm really lucky. I've been part
00:02:44
of some incredible teams and boards and
00:02:46
um and that's that's really uh been a
00:02:48
fun journey and I'm really excited to
00:02:50
share some of it with you.
00:02:51
>> So I mean just because Americus went to
00:02:53
that lecture but I didn't. So like
00:02:54
there's like a bunch of stuff that I
00:02:56
don't even understand the words of. So
00:02:58
can so I understand growth loop which is
00:03:01
the company you're CEO of uh has
00:03:04
something to do with compound marketing
00:03:06
like what what does that term mean and
00:03:08
and what does aentic AI mean? Can you
00:03:11
just start at the really basic?
00:03:13
>> Yes. Yes.
00:03:14
>> Cuz we know compound we know compound
00:03:16
interest is a good thing, Chris.
00:03:18
>> 100%. And that's where it started. You
00:03:21
know, in that small town, we used to get
00:03:23
a lot of snow in the winter. And I
00:03:24
remember my we'd have parents, friends
00:03:26
of our of ours come in and sometimes
00:03:28
they get stuck. Like literally, it' be
00:03:29
that much snow. And I was given a book
00:03:31
during one of these four-day stays when
00:03:33
some some family members read some books
00:03:35
and they left me this book. And it was
00:03:36
called The Wealthy Barber. And it was a
00:03:38
story about a proverbial wealthy barber
00:03:40
who did some really small investing and
00:03:42
compounded it over time and he became
00:03:44
wealthy. And that started a lifelong
00:03:46
fascination with the notion of
00:03:48
compounding. And when you stop and think
00:03:50
about it, compounding is everywhere.
00:03:52
Albert Einstein famously said it's the
00:03:53
eighth wonder of the world. Compound
00:03:54
interest, right? It's in nature, right?
00:03:57
If you think about how forests happen,
00:03:58
it's actually compounding. There's like
00:04:00
the worldwide wood uh worldwide wood
00:04:02
they call it. It's a mitochondria like
00:04:05
underneath the soil. It's in nature.
00:04:07
It's in technology. You know, when you
00:04:09
think about network effects, that's
00:04:10
compounding. Um James Clear, who I'm a
00:04:13
huge fan of, if you get 1% better every
00:04:15
day, by the end of the year, you'll be
00:04:17
37 times better. So, it's the notion of
00:04:21
little things that are done consistently
00:04:23
over time that start to lead to
00:04:25
exponential results. And we got to
00:04:26
thinking, okay, gosh, we're trying to
00:04:28
combine data and AI with people in the
00:04:31
loop to unleash the brand's potential,
00:04:33
right? whether it's a personal
00:04:34
connection with a brand or whether it's
00:04:36
the different variety of of tools and
00:04:39
channels you can bring together to offer
00:04:40
truly personalized and really great
00:04:42
experiences. So we borrowed the term
00:04:44
compounding shamelessly and say hey what
00:04:46
if we could compound marketing and so
00:04:48
that's really where it starts. Um the
00:04:51
second part with with aantic really like
00:04:54
AI is interesting in so many ways and
00:04:56
we'll unpack some of it today but many
00:04:58
of the changes in marketing were were
00:05:00
kind of the demand side meaning we
00:05:02
changed the consumers changed how they
00:05:04
interacted right we went from desktops
00:05:06
to mobile right analog to digital and
00:05:09
certainly there's elements of demand
00:05:11
side change here but the bigger change
00:05:13
is supply side what I mean by that is
00:05:15
the way in which we do marketing the way
00:05:17
in which we go about workflows uh with
00:05:20
AI in the middle uh is really exciting
00:05:24
and we're just scratching the surface
00:05:25
and again I'm keen to get into it with
00:05:27
you today because I think it's a supply
00:05:29
side meaning it's changing the way in
00:05:32
which we interact with people in a very
00:05:33
personal way in a way which in many ways
00:05:36
has been the holy grail forever and
00:05:38
we've been living with this unfulfilled
00:05:40
promise of marketing as a cost center or
00:05:42
marketing as some interruption and
00:05:44
there's so much more we can do together
00:05:46
and I know you feel the same uh and I'm
00:05:48
keen to as I said get into it a little
00:05:50
bit with you today.
00:05:51
>> So, okay. So, you had this big idea of
00:05:54
compound marketing and you understand
00:05:56
more of how all this works. So, with
00:05:58
that in mind, did you start growth loop
00:06:01
or like how did the whole company get
00:06:03
started or what is what is it exactly?
00:06:05
>> So, I've been fortunate remember I
00:06:07
talked about great brands and teams and
00:06:09
Google was one of them. I was at uh
00:06:10
Google for 10 years during a very great
00:06:13
in you know really really wonderful time
00:06:15
during a very obviously generational
00:06:17
company and during that time I met some
00:06:19
incredible people including the two
00:06:21
founders of growth loop uh they worked
00:06:23
for an old CMO of mine when I was up
00:06:25
running Google in Canada uh and they are
00:06:27
like just incredible entrepreneurs and
00:06:30
they they looked at how Google did
00:06:32
marketing and they said gosh for a
00:06:33
company that's allegedly so innovative
00:06:35
they sure don't don't like really always
00:06:37
think about marketing an interesting way
00:06:39
So Chris and David, the founders, said,
00:06:41
"Hey, why is it that I, as a product
00:06:43
marketing manager with goals to drive
00:06:46
value for a particular product, have to
00:06:49
go and line up outside a data team just
00:06:51
to basically get a hypothesis tested
00:06:53
with SQL and all this code and that
00:06:55
bounces back and forth and takes days
00:06:56
and weeks and up to a month just to do
00:06:58
one campaign. They thought that was in
00:07:00
that was silly, right? So basically they
00:07:02
had the innovation to to build on top of
00:07:05
a data cloud, right? So increasingly
00:07:07
we're putting data in fewer places these
00:07:09
things called data clouds that's Google
00:07:11
bigquery this is Google cloud company
00:07:13
like snowflake data bricks Azure um
00:07:17
Amazon there are many data clouds but
00:07:19
the notion was how about you start there
00:07:21
by putting an intelligence layer on top
00:07:24
of the data as opposed to pushing the
00:07:26
data everywhere really and this is
00:07:28
incredibly important from a cost
00:07:29
perspective from a compute perspective
00:07:32
security perspective and now of course
00:07:34
with AI you want to have your data and
00:07:37
your understanding and your intelligence
00:07:39
on your customers in one place so you
00:07:41
can apply the full effect of this
00:07:44
magical technology called AI. So that's
00:07:46
the insight where it started. And then
00:07:48
the other part with compounding is the
00:07:50
notion that it's not a funnel so much as
00:07:52
it's a loop, right? You basically have
00:07:54
ideas who you should be speaking to,
00:07:57
where you should be speaking to them and
00:07:58
then do closed loop experiments where
00:08:00
you get a little bit better every time
00:08:02
and then it really compounds as you do.
00:08:04
So
00:08:04
>> wait, I got you got a because I'm like
00:08:07
big on this funnel and this loop. So I I
00:08:10
really need you to unpack that a little
00:08:11
more. So you have a different structural
00:08:14
model of dealing with the customer
00:08:16
rather than thinking about it like
00:08:19
>> a funnel which is traditional. So you
00:08:21
know are you in the product category
00:08:23
then you're considering the brand then
00:08:25
you know my brand compared to others
00:08:27
you're suggesting some kind of loop. So
00:08:29
what exactly does that mean? I I I don't
00:08:32
really understand.
00:08:33
>> Yeah. No great question. Like so in a
00:08:35
funnel, we basically treated the world
00:08:37
as like this big uh this big pool and
00:08:40
then you basically pull them through a
00:08:41
funnel. Um and there's a consideration
00:08:43
all that stuff we've been we taught when
00:08:45
I was in business school. We were taught
00:08:46
about consideration, awareness, all that
00:08:48
good stuff. And look, I'm here to say
00:08:50
look, you need to continue to invest in
00:08:52
your purpose and your brand and why you
00:08:54
why you exist. That is that is as
00:08:56
important if not more important than
00:08:58
ever. However, you can do better, right?
00:09:01
Rather than treating everyone as like
00:09:02
one big segment, we're actually getting
00:09:05
close to the holy grail where you can
00:09:06
actually reasonably into it something
00:09:09
interesting about a person at the
00:09:11
onetoone level. This is the importance
00:09:13
of what sometimes called first party
00:09:15
data information you have on that
00:09:17
person. Okay. Um so that you can inter
00:09:20
you can intercept them at the at the
00:09:22
right moment
00:09:23
>> to ensure loyalty surprise and delight
00:09:26
them. the customer journey kind of model
00:09:27
then
00:09:28
>> it's an iterative customer journey that
00:09:30
says you're going to treat people or sub
00:09:32
subsegments or even onetoone for the
00:09:35
first time in an iterative loop as
00:09:37
opposed to just doing what I sometimes
00:09:39
call waterfall marketing.
00:09:40
>> It's more agile just like we develop
00:09:43
software.
00:09:44
>> Yeah. Just let me let me just see if I
00:09:46
got if I have this correct, Chris,
00:09:48
because I I don't think what you're
00:09:49
saying is somehow
00:09:51
>> uh falsifying the funnel as much as it's
00:09:54
saying, hey, you know what? We can make
00:09:56
this funnel super efficient by almost
00:09:58
pointing and using tools that allow us
00:10:01
>> to kind of get to that that behavioral
00:10:03
part faster. In other words, we can have
00:10:06
some understanding and then we can use
00:10:08
artificial intelligence to sort of test
00:10:11
kind of immediate sort of intuitions
00:10:13
that we have and quote get to the right
00:10:15
answer as fast as possible. So there's
00:10:17
less guesswork I guess in this
00:10:20
>> also predicated on this notion of touch
00:10:22
points like I mean how do you know when
00:10:24
to interact? It has to be when the
00:10:26
customer engages I guess.
00:10:29
>> Exactly. Exactly. So, quick story. I
00:10:31
serve on the board with a very senior um
00:10:33
member at Starbucks, executive at
00:10:35
Starbucks, rather, and he he's he he and
00:10:37
I were talking about this this very
00:10:38
topic, and he says, "Look, we know that
00:10:41
uh Sally is a vegetarian. So, why is it
00:10:44
when Sally visits her local Starbucks,
00:10:47
we're offering her meat products? We're
00:10:50
meat." Like, that just seems odd, right?
00:10:52
when you can reasonably know with
00:10:54
permission and Sally wants to know like
00:10:56
what she she's gonna get when she greet
00:10:58
is greeted by the barista they probably
00:11:00
know that but systematically how about
00:11:03
you offer something that would be
00:11:04
relevant to Sally as opposed to relevant
00:11:06
so it does start absolutely touch points
00:11:09
a great a great point
00:11:11
>> so when Sally comes to the cash register
00:11:14
>> yeah you learn about
00:11:15
>> a script about Sally that's what you're
00:11:17
saying right
00:11:19
>> exactly that's customer engages with us.
00:11:22
Is that what you're saying?
00:11:23
>> Yeah.
00:11:24
>> Exactly. Exactly. And that's just one
00:11:25
touch point. I love that term. We
00:11:26
actually, you know, we talk about touch
00:11:28
points and moments that matter. Uh
00:11:31
because again, sometimes it's it's
00:11:34
leaving Sally alone, right? It's not
00:11:36
just constantly interrupting. It's
00:11:37
>> implies you know Sally and Sally's going
00:11:39
to get to know you, right? I mean, so
00:11:41
are you only talking about a situation
00:11:43
where I've already had interaction with
00:11:45
the customer and then the way I interact
00:11:48
with her at the next moment should
00:11:50
reflect that history. But that implies I
00:11:53
have that relationship already.
00:11:55
>> Correct. That's correct. Right. So what
00:11:57
we but I am talking about both, right?
00:11:59
So you can reasonably intuit it. And
00:12:00
this is where back to machine learning,
00:12:02
right? So we we sometimes intermix the
00:12:03
term machine learning with AI. Machine
00:12:05
learning says, okay, how can we go about
00:12:08
into the rest of the the potential
00:12:11
prospects or potential touch points to
00:12:13
basically inform the set of next actions
00:12:16
which would reasonably do do the right
00:12:19
thing for the brand and for the
00:12:21
long-term customer relationship. And
00:12:23
this is a really exciting part where you
00:12:25
can actually get to causality, right? So
00:12:28
most of marketing is correlative. It's
00:12:29
like we think these things worked but we
00:12:32
don't know, right? And that's the thing
00:12:33
that's holding back a lot of marketing
00:12:36
is like, you know, BF know the CFO and
00:12:38
the CMO need to be BFFs, right? It needs
00:12:41
to be you talk the language of the CFO
00:12:44
and that's really one of the things that
00:12:46
I think separates the great CMOs from
00:12:48
the sort of mediocre ones.
00:12:50
>> Yeah, that's I love the CFO and the CMO
00:12:53
need to be BFFs. That's fantastic. I
00:12:55
need that on a t-shirt, Chris. All
00:12:57
right. Now, listen. So I I want to make
00:12:59
sure I'm I'm unpacking this correctly
00:13:01
because I think what you're saying is
00:13:02
like hey the fact that we have tools
00:13:04
that allow us to unpack insights and
00:13:06
data very quickly now allows us to be
00:13:09
much more
00:13:11
>> uh much more inductive right so we can
00:13:13
kind of go to Barbara's point we can
00:13:14
kind of immediately we have a touch
00:13:16
point with Sally and we have we have
00:13:18
some inferences about what she's done in
00:13:20
the marketplace and not what she
00:13:21
reported on a survey but what she's
00:13:23
actually done and we start building that
00:13:25
intelligence and testing and iter
00:13:27
Iterating and testing and iterating and
00:13:29
testing to get that clear picture of
00:13:32
Sally and then see to what extent can we
00:13:35
uh predict other Sally's out there as a
00:13:38
function of is that close at all in
00:13:40
terms of
00:13:40
>> you know so fast personalized data
00:13:44
driven and continuously optimized. This
00:13:46
is the notion of a funnel and this is
00:13:48
exactly what we're talking about. So you
00:13:49
absolutely nailed it. But so his point
00:13:52
though because I I understand once you
00:13:54
know the person using that data that
00:13:56
makes sense to me but it I what America
00:14:00
said which is taking it another level.
00:14:02
And is that what you're saying? The way
00:14:03
you reach people you don't know
00:14:06
>> is by optimizing the data you do know.
00:14:08
So that's where you start and then you
00:14:10
look for other people like that's what
00:14:12
America said. Is that what you're saying
00:14:13
too?
00:14:14
>> Yeah. Correct. So you're absolutely
00:14:15
right. There's two modalities. Do you
00:14:17
know the person? Do you know something
00:14:19
about them or do you not? And that's
00:14:20
like the different techniques are used
00:14:22
to to tease those apart. But where it
00:14:24
starts to get magical is when you start
00:14:26
to for inform uh one from the other. Um
00:14:29
a lot of the a lot of the science is b
00:14:31
sort of the techniques are born from a
00:14:33
lot of probabilistic statistics, right?
00:14:35
So you'll you'll hear things like
00:14:36
multi-armmed bandits and all that's a
00:14:38
fancy way of saying is you're assigning
00:14:40
people to a group and then you have a
00:14:42
test and control, right? A lot of the
00:14:44
power and things we're super excited
00:14:46
about is that you actually have a test
00:14:48
and control group. So you can prove
00:14:50
causality. You can basically and have a
00:14:51
hold out group that lasts for a time.
00:14:53
Now here's an interesting magic we're
00:14:55
we're really excited about is over time
00:14:57
you can use math to basically
00:14:59
approximate the control group. So you
00:15:01
don't actually have to hold them out and
00:15:02
still determine causality and still
00:15:05
determine is it having the lift and the
00:15:07
desired personalized impact and moving
00:15:10
the needle on a metric like your
00:15:12
customer lifetime value. That's like
00:15:14
whoa. That's like mind-blowing for many
00:15:16
people and we're really just scratching
00:15:18
the surface on what's possible there.
00:15:20
>> Interesting. Let me ask this question. I
00:15:21
love this point because in some senses
00:15:23
I've heard this quote, Barbara, that
00:15:26
marketing starts with math and ends with
00:15:28
art. Uh kind of I don't we'll unpack
00:15:32
that later, Barbara. But Chris, help me
00:15:36
because I want to get to to Barbara's
00:15:37
level of of understanding here. very
00:15:40
specific example to say all right I come
00:15:42
to you Chris in growth loop and I've got
00:15:44
a data set
00:15:45
>> is that how this happens and then and
00:15:47
then you you look at my data
00:15:50
>> and then you build out a kind of AI
00:15:53
infrastructure andor agentic
00:15:55
architecture to lay on top of my data in
00:15:59
order to help me pull as much insights
00:16:01
about what I can understand from the
00:16:03
people in that data set as fast as
00:16:05
possible is that some is that correct at
00:16:07
all correct let's make it let's make Get
00:16:09
specific. Okay. So, a brand comes to
00:16:11
usific.
00:16:12
>> Yeah. Yeah. So, so they have a brief,
00:16:13
right? We're familiar with marketing
00:16:14
briefs. Just say this is what we were
00:16:16
trying to accomplish is the business
00:16:17
objectives is who we're trying to talk
00:16:18
to, right? Literally drag and drop the
00:16:20
brief into our agent, our campaign
00:16:23
agent, right? We basically start the
00:16:25
following. Say discerning what the what
00:16:28
the market is trying to accomplish. And
00:16:30
you talk about art. I think art and and
00:16:32
and math intertwine at different stages
00:16:35
because it's like the the real role for
00:16:37
humans is ingenuity and creativity to
00:16:39
say hey what's happening in this in the
00:16:41
zeitgeist that really is relevant to our
00:16:43
customers. So it's a really important
00:16:44
role marry that up with what has worked
00:16:46
in the past. So a brief comes in it
00:16:49
basically gets processed by an agent and
00:16:50
it says what do we know about this um
00:16:54
this objective and these set of
00:16:55
campaigns we've run in the past and the
00:16:57
people that might be relevant. So it
00:16:58
then it then offers up to the marketer a
00:17:01
set of different audiences to basically
00:17:03
go and and and you know treat with
00:17:06
something and then it would suggest a
00:17:08
journey right to say okay maybe you send
00:17:10
them uh something through your app or
00:17:12
your email or you know may maybe you
00:17:14
have to give them some time and maybe
00:17:15
then offer them an ad or some form of
00:17:17
incentive. So that's that's what
00:17:20
happens, right? A brief gets turned into
00:17:22
a specific set of audiences and a
00:17:24
specific journey which then gets pushed
00:17:26
into the world and then executed and
00:17:28
then run back into what worked and read
00:17:31
it back into your data cloud. Lather,
00:17:33
rinse and repeat. This is exactly the
00:17:35
work that we're doing with some
00:17:37
>> Can you give me like a real concrete
00:17:39
example with specific like just
00:17:41
specifics exactly how this worked like a
00:17:43
big success story?
00:17:45
>> Sure. I I'll give you a recent one. We
00:17:47
we are so fortunate to work with Costco,
00:17:49
right? Everyone knows and loves Costco.
00:17:52
>> Chris, you're speaking her language.
00:17:54
COSTCO
00:17:56
ALREADY.
00:17:58
>> I love Costco. So, um I'm going to give
00:18:00
a little bit of a little spin on this,
00:18:02
but Costco is super super into their
00:18:05
members, right? And they they put their
00:18:07
members above all else. Their business
00:18:08
model is well predicated on selling
00:18:10
memberships and they're very tight. Like
00:18:12
if you ever um you know what the
00:18:13
acquired podcast has done a deep deep
00:18:15
dive
00:18:15
>> into I've listened to that several
00:18:17
times. It's great
00:18:18
>> and I love it. So okay so people are
00:18:20
familiar with Costco and and Costco is
00:18:22
very in the in the business of working
00:18:24
with their brands to basically ensure
00:18:26
delightful experiences and connecting
00:18:28
them to you know say it's DeWalt tools
00:18:30
right so basically how can you infer the
00:18:33
people in the Costco membership that
00:18:35
might have a propensity to buy DeWalt
00:18:37
tools we can do that with math right and
00:18:40
then what we're doing is connecting the
00:18:42
two we basically say DeWalt Costco here
00:18:45
is here's a subset of customers that
00:18:48
used to take you days and weeks just to
00:18:50
do to to to really do the math and
00:18:53
really surface the people. We now do
00:18:55
that in hours, right? In terms of really
00:18:59
understanding and automating the
00:19:01
connection between in this case a brand
00:19:04
and a set of high propensity customers.
00:19:06
>> What do you do with that information?
00:19:07
Can you show me like
00:19:09
>> Yeah. Then then you do you send them an
00:19:11
email? Do you send them an offer? Do you
00:19:12
push something into the the Costco app?
00:19:14
It really depends. and you're constantly
00:19:16
testing to understand both what offer
00:19:18
resonates with the subset, which
00:19:20
channels offer the off um seem to work
00:19:22
the best. Maybe it's an instore um
00:19:25
experience. So, it's it's a really
00:19:26
multi- channelannel or omni channel.
00:19:28
Sometimes people use that term
00:19:30
experience that says, okay, how do we
00:19:31
broker this in a way that is additive
00:19:34
for the merchant in the home improvement
00:19:36
category and of course the members? It
00:19:38
has to be a balance between the two.
00:19:40
>> So, so let me just get the it's just
00:19:42
hard for me. So your client is DeWalt
00:19:46
Tools and they want to know
00:19:48
>> or or your client is Costco and so
00:19:51
Costco pays you to figure out how to
00:19:54
optimize the purchases of specific
00:19:56
brands within their environment. Is that
00:19:58
how it works or
00:19:59
>> that that is one use case exactly?
00:20:01
Right. So it is and really just to
00:20:04
abstract it away and talk just in terms
00:20:06
like it allows their marketing velocity
00:20:08
to be faster. Right. Like where that
00:20:09
used to take weeks, it now takes hours.
00:20:11
So therefore, right, DeWalt has money
00:20:14
they want to spend and there's many
00:20:15
DeWalts, right? There's n number of
00:20:17
brands
00:20:18
>> and if it's very manual and takes a long
00:20:21
time.
00:20:21
>> Yeah.
00:20:22
>> Well, so this is like a retail media
00:20:24
network. Is that who your clients are?
00:20:26
>> In this case, that's what it is.
00:20:27
Correct. Correct. But we do we do life
00:20:29
cycle marketing and retail media
00:20:31
networks and uh really we do the similar
00:20:33
stuff with Albertson's and you know
00:20:35
price line and many others.
00:20:36
>> Okay. Okay. So your client is the
00:20:39
retailer and the retailer's customers
00:20:42
are these brands and you're trying to
00:20:44
figure out a way to optimize what the
00:20:47
brand can get from the retail media's
00:20:49
data and the way they optimize all that
00:20:52
is through you. Is that right?
00:20:54
>> Correct. Correct. We we play we play a
00:20:56
part in a broader ecosystem. I want to
00:20:58
be clear. But yeah, it's it's it's the
00:21:00
how do you accelerate that get the
00:21:02
velocity? We talk about speed. Yeah. And
00:21:04
then secondly, did it work? Right. It's
00:21:07
not just did you did you throw out some
00:21:09
impressions. No, no, no. Like the the
00:21:11
brands with all respect, they care about
00:21:12
moving their product and having dectual
00:21:14
experiences. So, we also help them
00:21:15
measure incrementality,
00:21:17
>> right? Right.
00:21:17
>> In the store, right? And that's that's
00:21:19
sounds easier than it is in terms of the
00:21:21
actual math and the calculation. But
00:21:23
>> no, incrementality defined as if you
00:21:26
didn't do anything, how much more with
00:21:29
what you did matters in some sense.
00:21:31
Bingo, right?
00:21:32
>> Yeah. It's a resource allocation problem
00:21:34
in some senses. Let me ask you this,
00:21:36
Chris, because I I want to make sure I
00:21:39
understand this part because the math
00:21:40
gets the retailer to some hypothetical
00:21:44
uh set of audiences to think about.
00:21:47
>> Does the retailer then step in because
00:21:49
now it's sort of like, okay, what do I
00:21:50
prioritize? Is it an ad? Is it a this?
00:21:53
Is it is it a that? Is that where the
00:21:55
marketers then step in and say, "Well, I
00:21:56
know something about some of these
00:21:58
groups and now I'm going to bring sort
00:22:00
of a creative piece to sort of
00:22:01
prioritize some of the marketing uh
00:22:04
protocols that might be used to again is
00:22:07
that what happens in this case?"
00:22:09
>> Yeah. Exact. We believe in human in the
00:22:10
loop, right? So these agents are working
00:22:12
on behalf of the brands and the
00:22:14
customers and they can suggest things
00:22:16
the audience the journey the creative
00:22:18
creative like Nano Banana and all these
00:22:20
images are getting like so good, right?
00:22:22
So I wouldn't have expected that uh the
00:22:24
pace we can we can talk about that
00:22:26
separately but yeah it's about
00:22:27
suggesting things and yeah it's using
00:22:28
intuition right it's using the
00:22:30
understanding of what has worked in the
00:22:32
past so so marketers arguably they're
00:22:34
getting out of the mundane stuff right
00:22:35
and a lot of that you know the work of
00:22:37
the work is so not why marketers signed
00:22:39
up for the job so we're freeing them up
00:22:41
to do the job that maybe they wanted to
00:22:43
do in the first place
00:22:44
>> along the lines of what America is
00:22:46
asking are a lot of the suggestions more
00:22:48
price related or deal related or I mean
00:22:51
typically Great question. Yeah. Um, so
00:22:54
you know this is where machine learning
00:22:56
and AI actually interact. So machine
00:22:58
learning would be a propensity model. So
00:23:00
we have high propensity like take
00:23:02
Albertson's. Albertson's knows who's a
00:23:04
vegan, right? So they have like
00:23:06
propensity models to say vegans uh have
00:23:09
high propensity to provide these
00:23:10
products, right? These products and not
00:23:12
these ones. Uh or to your point Barbara,
00:23:15
price elasticity, right? Um I serve on
00:23:17
the board of GAP and tariffs have had
00:23:20
enormous impact on anyone who's
00:23:22
importing things and GAP certainly falls
00:23:23
into that camp.
00:23:24
>> So getting an understanding of where and
00:23:27
how you can change prices that's a price
00:23:29
elasticity game right. So how do you
00:23:31
take
00:23:32
>> that would be wonderful.
00:23:34
>> I mean it's it's f it's a hard problem
00:23:36
but it's like not just not just in a
00:23:38
category all the way down to a skew
00:23:39
level and then ideally at an individual
00:23:41
level. So this is where the the math
00:23:43
those are those are math problems right
00:23:45
machine learning like propensity or
00:23:48
price elasticity.
00:23:49
>> Yeah. So, let me let me just go back to
00:23:51
the old old world of grocery. Like
00:23:53
before we had all this data and all this
00:23:55
fancy stuff and you can do it on each
00:23:57
customer, whatever. I remember Catalina
00:24:00
would would put out coupons in the
00:24:02
grocery based on what you purchased just
00:24:05
at the cash register that time because
00:24:06
they didn't have all the sophistication.
00:24:08
And they do one thing or another, which
00:24:10
is kind of what I hear you saying.
00:24:12
They'd either give me a price deal of
00:24:14
some kind, some kind of coupon for
00:24:17
future purchase, or if I was Pepsi and
00:24:21
paying for it and the customer bought
00:24:23
Coke, they might give me a coupon for
00:24:26
Pepsi to try to convert me. So, like
00:24:28
what you're saying now is in
00:24:30
multi-dimensions, much more in a
00:24:32
sophisticated way than Catalina used to
00:24:34
do it at the one time at the cash
00:24:36
register. I can either give you
00:24:39
different kinds of price deals or like
00:24:41
you said if I'm vegan or I can give you
00:24:43
a constellation of brands where I'd be
00:24:45
more that's what you meant by propensity
00:24:47
I think more likely to buy that product
00:24:51
because I see you bought this in the
00:24:53
past therefore you're more likely to buy
00:24:55
carrots than you are to buy steak
00:24:59
>> with the veganism is that what is that
00:25:01
kind of very I'm saying in super simple
00:25:04
way and you're doing it in a
00:25:05
multi-dimensional way but is and 10
00:25:07
times as fast probably. Well 10
00:25:12
that is well said Barbara exactly I
00:25:13
remember Catalina well you studied him
00:25:15
and that's exactly now it's it's a it's
00:25:16
a version of that but it's like it's
00:25:18
it's so it's faster to your point and is
00:25:21
much more much more it take is taking in
00:25:23
so much more data right that's a very
00:25:25
simple if this
00:25:26
>> is a very simple example but I can
00:25:28
understand it but now if you multiply it
00:25:31
by all these different directions and do
00:25:33
it like instantly the power is
00:25:35
incredible
00:25:37
>> it is it is and you know so let's weave
00:25:38
in you asked early off the top, aantic,
00:25:40
what does that mean? Like the reason,
00:25:42
you know, I spent three hours yesterday
00:25:44
doing a cloud code um course for PMs in
00:25:47
this case, but like I I so recommend
00:25:49
people jumping in and trying it and and
00:25:51
why do I get excited? So the the the
00:25:53
innovation that cloud code and and
00:25:55
variants of uh of it represent or what
00:25:57
they're calling long horizon uh agents,
00:26:00
right? So you know when when uh the chat
00:26:03
pimo would happen that was all about
00:26:04
pre-training a model and then we had
00:26:07
reasoning and inference time compute
00:26:08
like that was where like it's thinking
00:26:10
to basically really really run lots of
00:26:12
different permutations and this long
00:26:14
horizon agents are a gamecher because it
00:26:16
actually can look long horizon and hold
00:26:19
context and look at the whole system. So
00:26:21
to your point, Barbara, this is not only
00:26:24
just taking in more information. It's
00:26:25
actually thinking about the long horizon
00:26:27
of the relationship, right? So we're
00:26:29
just starting to see this where it's
00:26:30
like, well, maybe maybe it is like these
00:26:33
se these 10 things. I'm making that
00:26:35
number up, but it's thinking about the
00:26:37
entire system and that's where the agent
00:26:40
it is.
00:26:41
>> So you're maximizing lifetime value.
00:26:43
You're not only at this one occasion
00:26:45
doing it in a million different
00:26:46
directions. So, it's not only
00:26:48
multi-dimensional in that way with the
00:26:50
different tools you can use and all the
00:26:52
different questions you can ask, but
00:26:53
it's dynamic. It's long term. So, you're
00:26:55
looking about how to maximize the the
00:26:58
That's incredible.
00:27:01
>> Ladies and gentlemen, Chris O'Neal has
00:27:03
just split the atom slide on our
00:27:06
podcast.
00:27:07
>> I mean, amazing. I I can't I mean, I see
00:27:10
why it's so hard to explain because it's
00:27:12
just so much. Then you have metrics that
00:27:16
prove you're able to do this somehow or
00:27:17
another.
00:27:18
>> I mean, you're not just taking it on
00:27:20
faith cuz it's pretty hard to get your
00:27:22
brain around all of these things.
00:27:24
>> It it is, but you know, you know, the
00:27:25
the math the math we've applied it in ad
00:27:28
tech for a long time. Those previous,
00:27:31
you know, run it Google and now this is
00:27:33
like how do you bring it into marketing
00:27:34
personalization. So, it's borrowing
00:27:36
similar concepts and then and then
00:27:37
turbocharging with AI. It's just like
00:27:39
it's it is my my bending. You know the
00:27:42
hardest thing is actually changing the
00:27:44
people's minds and the workflows in
00:27:46
their life. People get used to different
00:27:48
ways of thinking and the longest pull
00:27:50
the hardest change is actually changing
00:27:52
the ways in which people go about their
00:27:54
business and that's the hardest thing
00:27:55
and that's you know I find it deeply
00:27:57
rewarding once you do that. So you're
00:27:58
not just applying your existing
00:28:00
workflows and your mindsets to a whole
00:28:02
new set of tools or vice versa right new
00:28:04
tools on top. When you say people, are
00:28:06
you talking about the marketers, the end
00:28:07
users, the retailers? Who are the people
00:28:09
you're talking about?
00:28:10
>> All the all the above, but particularly
00:28:12
the marketers, right? And so like what's
00:28:14
happened? There's always been a
00:28:14
tugof-war between technical people and
00:28:17
like the data teams and lots of
00:28:19
different parts of large organizations.
00:28:21
>> You know, if you'd asked me 10 years
00:28:22
ago, I would have thought that the the
00:28:24
marketing leaders of today or now would
00:28:27
be a blend of art and science, right?
00:28:29
These people who are have data driven
00:28:31
thing and and deep appreciation for
00:28:33
personalization and brand. I would have
00:28:34
thought they sort of merged those
00:28:35
together and that's still largely true.
00:28:38
But what's happening is the data teams
00:28:40
and the technologists are really leaning
00:28:42
in and they're saying, "Hey, you know,
00:28:43
for a long time this promise has not
00:28:45
been kept, right? The the BFF thing I
00:28:47
mentioned." The reason that's not true
00:28:50
as a rule is that basically marketers
00:28:53
haven't leaned in consistently to show
00:28:55
that this isn't just a cost center. This
00:28:57
is a growth engine. Right? So the
00:28:58
technologist with all these tools we're
00:29:00
talking about, I mean, they're so
00:29:01
powerful, it's unbelievable. I wouldn't
00:29:03
have the slope of improvement is
00:29:06
incredible. So you're seeing the
00:29:08
technologists really play a role and
00:29:10
they're trying to learn marketing faster
00:29:12
than marketing is trying to learn tech
00:29:13
and it's a fascinating time to be in
00:29:15
business and but it does require a
00:29:17
mindset and a partnership of
00:29:18
collaboration that takes time and you
00:29:21
know really really um the culture.
00:29:23
>> So let's you know just we we are almost
00:29:26
at the end of time if we're not at the
00:29:27
end of time. So let's just big picture
00:29:29
this like just to get like
00:29:31
>> with all of this power and all this
00:29:33
ability that you have to make all these
00:29:35
predictions you know are you would you
00:29:38
maintain that you are not only
00:29:41
increasing
00:29:43
bottom line for the retailer because the
00:29:45
retailer sending this information so the
00:29:47
more useful the information is the more
00:29:49
money they'll get for that information
00:29:50
the more value they'll have the marketer
00:29:53
who's selling more product but are you
00:29:55
also maximizing the experience for the
00:29:58
end user like is that one of the goals
00:30:00
too? I mean,
00:30:01
>> I think it's the main goal. Absolutely.
00:30:03
Right. For too long, I mean, the reason
00:30:05
I'm in this line of work is like, look,
00:30:07
we all experience this in our day-to-day
00:30:09
life, some sub-optimal experiences,
00:30:10
whether it's the actual product
00:30:11
experience or the marketing experience,
00:30:13
the customer service experience. And
00:30:14
like I just think that, you know, this
00:30:16
is one of going to be one of the use
00:30:17
cases just like it is for DevOps
00:30:19
developing code or customer service
00:30:22
interaction where you're applying
00:30:23
agents. Same as in marketing, too. For
00:30:25
goodness sake, we know we know
00:30:27
information and people are willing to
00:30:29
offer it. in a fair exchange for more
00:30:32
delightful experiences and that's what
00:30:34
right great brands I know I think I'm
00:30:35
preaching to the to the choir here that
00:30:37
like that's what it's all about that's
00:30:39
the higher order bit you know the tech
00:30:40
can be a service to wonderful
00:30:42
experiences
00:30:43
>> and that's what um that's what gets us
00:30:45
out of bed every morning and we're
00:30:47
really really thrilled to put gonna put
00:30:50
up for me next like that's I'm going to
00:30:54
be even happier shopping at Costco now
00:30:57
>> that's funny I love it though Barbara to
00:30:58
your point though because as Chris
00:31:00
saying it's the trifecta, right? I mean,
00:31:02
consumer welfare goes up, marketing
00:31:04
efficiency, marketers do their jobs
00:31:06
better. I mean, every it's a just
00:31:07
win-winwin. This is fantastic.
00:31:10
>> Well, Chris, it's been kind of an
00:31:12
amazing little session. If I understand
00:31:14
an eighth of what you're talking about,
00:31:16
I'm thrilled. But, uh, it's been
00:31:19
wonderful talking to you. Can you tell
00:31:20
us where consumers can learn more about
00:31:22
this or our listeners can learn more
00:31:24
about what you're talking about and get
00:31:26
into more of the details? you have is
00:31:28
that available for people to dig in?
00:31:30
>> Yeah. Yeah, people can track me down on
00:31:32
LinkedIn. That's usually where I spend
00:31:33
spend my uh my my social time. Um so
00:31:37
people want to track me down on LinkedIn
00:31:38
or check out Growth Loop. Um we'd be
00:31:41
happy to to chat or or just really like
00:31:43
help us move this this whole uh industry
00:31:45
one one step forward. So thank you so
00:31:48
much for for the time and discussion.
00:31:49
I've enjoyed it very much. So thank you
00:31:51
very much for having me here.
00:31:52
>> Thank you. Chris O'Neal, CEO of Growth
00:31:55
Loop. That's all we have time for today.
00:31:57
We'd like to thank our producers, Dion
00:31:59
Simpkins and Marissa Rena. Thank you all
00:32:02
for listening. We'll be back next week.
00:32:04
Till then, this has been Marketing
00:32:05
Matters on the Wharton Podcast Network.
00:32:08
I'm Barbara Khan here with Americas
00:32:11
Reed.
