July 19, 2023 • 26 mins
By now, most marketers have dabbled in AI and introduced AI-powered tools to their creative workflows. In our last two episodes, we discussed how marketers are using these tools to work smarter and faster. But one area we haven't talked about is data -- an exciting opportunity for marketers.
Joining me this week are two members of the Plus Company team – Kristin Wozniak, Global Lead of Customer Success and Michael Cohen, Global Chief Data & Analytics Officer. We’ll go beyond ChatGPT and Midjourney to examine how Data and generative AI can improve marketing business operations and increase ROI.
We'll talk about how we as marketers can leverage AI to better allocate targeted ad spends. We’ll define a composable CDP, and how it and data-centric AI will change the way Marketers work.
We’ll also try to imagine how the future with AI will look for Marketers, recognizing it's nearly impossible to predict the unpredictable world of AI.
Learn more about Chip War, the book Brett mentions: https://www.christophermiller.net/semiconductors-1
Learn more about the composable CDP:
https://a16z.com/2023/05/22/the-rise-of-composable-the-cdp/
from plus company
Joining me this week are two members of the Plus Company team – Kristin Wozniak, Global Lead of Customer Success and Michael Cohen, Global Chief Data & Analytics Officer. We’ll go beyond ChatGPT and Midjourney to examine how Data and generative AI can improve marketing business operations and increase ROI.
We'll talk about how we as marketers can leverage AI to better allocate targeted ad spends. We’ll define a composable CDP, and how it and data-centric AI will change the way Marketers work.
We’ll also try to imagine how the future with AI will look for Marketers, recognizing it's nearly impossible to predict the unpredictable world of AI.
Learn more about Chip War, the book Brett mentions: https://www.christophermiller.net/semiconductors-1
Learn more about the composable CDP:
https://a16z.com/2023/05/22/the-rise-of-composable-the-cdp/
from plus company
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:01):
You think chat GPT would come upwith a really good answer and why it's
good for humanity You've given it's howimportant the future of humanity is to chat
GPT and its success. But anyways, I digressed. In a never changing
world, campaigns are all about collaborationand connection, from an initial brief to
(00:24):
the follow through, we explore howinnovative thought leadership makes a campaign success more
than a possibility. Welcome to partnersand Possibility. I'm your host, Brett
Marshaun, CEO of Plus Company.By now, most marketers have dabbled an
AI and introduced aipower tools to theircreative workflows. In our last two episodes,
we discussed how marketers are using thesetools to work smarter and faster.
(00:47):
But one area we haven't talked aboutis data and exciting opportunity for marketers.
Joining me in this episode are twomembers of the Plus Company, t Kristen
Masniak, SVP of Customers six andMichael Cohen, Global Chief Data and Analytics
Officer. We'll go beyond chat GPTand mid journey to examine how data and
generative AI can improve marketing, businessoperations and increase ROI. We'll talk about
(01:15):
how we as marketers can leverage AIto better allocate targeted AD spends. We'll
define a composable CDP and how itand data centric AI will change the way
marketers work. We'll also try toimagine how the future AI will look for
marketers, recognizing it's impossible to predictthe unpredictable world of AI. Michael,
(01:40):
you say something on a regular basis, and it's about the horse versus the
car. Where if you ask peopleto protect and what they want. You
know, in the early nineteen hundredsthey say a faster horse. No one
would have necessarily imagined the car.And so I think there's an element too
of thinking through what it is thatwe want from the future of AI,
and then what will surprise as usgoing forward. Fundamentally, I think it's
(02:02):
important that we keep an eye onour demands, so to speak, or
there'll be a collection of new opportunities, new solutions to meet our demands.
And my demands I don't mean likemy demands, but things that I need
help answering questions, decisions. It'llbe a question of things we don't even
know what they are yet because wecan't quite yet, but hopefully to answer
(02:24):
some of our questions. Yeah,it's I mean, that's an interesting analogy.
I was thinking. I'm reading thisbook about it's called chip Wars,
which I highly recommend. It talksabout sort of the history of the transistor
and and cilicon chips and what it'sdone and Moore's law, et cetera.
And you know, everyone just thoughtit was going to be used in the
(02:46):
beginning for the military, and itwas really when people looked at it beyond
that that they realized that it hadthis incredible power. And it's kind of
the same thing with AI, right, I often say to people, even
though everyone is thinking about generative AIas far as producing text and some people
are thinking about it, you know, around images, not a lot of
(03:07):
people are thinking about it when itcomes to good old fashioned marketing dollars.
And I don't know, for everydollar spent on content, there's a ten
dollars spent on distribution and media.So you know, when you think about
it that way, there's probably tentimes the opportunity when it comes to measurement
and management. How are you seeingit helping with measurement, with management and
with data outputs? You know,the big picture here is that this sort
(03:30):
of prediction. Generative a on whateverwe want to call it, produces information,
whether that's in the form of languageor audio, or image or numbers,
and data answers the questions that wehave allows us to do that in
an iterative way. You might ask, what was my reach for men in
the sixty five to seventy five yearold to age group in southern California,
(03:53):
and how did they respond to aparticular campaign? Like you can layer all
those demensions on it at once.It will synthesize an answer based on the
information it has. It's not requiredto have all the data, and the
reality is that data is imperfect,but prediction can overcome that. One of
(04:15):
the interesting things about what you justsaid, Michael, especially thinking about the
vast collection of clients and regions,is data availability is so mixed. The
generative AIP starts to make it possiblefor clients, for people who haven't been
able to engage in this space before, because they can work with imperfect data
in a more meaningful and reliable way. Right. You know, there's that
(04:38):
famous quote by John Wannamaker, whichis I know half my advertising is working,
I'm just not sure which half,And I think one of the interesting
things about AI and how it canbe used in a predictive way, as
you just both described, is ifpeople can focus on what's working and how
do we do more of that,and unless like is it working or not?
(04:58):
And do I trust it or not? Etc. Tell me a little
bit about how we're using or you'reusing AI now, in particular, not
on the generative side of content,but on this management and measurement side that
we're talking about. Really, thefundamental sort of building block of the of
the marketplace that we aim to influenceis the customer and understanding that entire communication
(05:23):
journey we're having with them. Weknow about all the challenges to filling out
that story, data that exists behindwalled gardens, privacy regulations, you know,
low quality telemetry, just missing pieces. But AI or generative AI can
can solve the puzzle and predict whatthat journey looks like. And now we
(05:46):
have a building block to be ableto um to tell all of these combined
stories about audience, creative media,timing, different KPIs the journey through the
funnel, and do so in away that allows us to build technologies that
answer the questions we have today basedon how mature we are as data and
(06:10):
AI sort of organizations, but alsoanswer the questions of the future that we
don't know because we've built it aroundthe customer, who's always going to be
the building block of the marketplace.I think there's an interesting conversation here around
journeys in the funnel specifically. I'vestruggled with the funnel of my whole career,
and not because you know what isit eighteen ninety eight. It was
(06:32):
developed and stuff has changed since then. But I think more and more as
the industry changes, the way wepurchase things change, the funnel isn't as
straightforward. I don't think we canthrow it away. There's lots of categories
and instances where the funnel and thatlinear journey does make a lot of sense,
and there's other areas now where itreally really doesn't. And if I
work with different clients, everybody comesto the table with their own sort of
(06:56):
linear version of what that journey lookslike. Even if it is a goal,
there's this one model, and theconversation then from the agency side specifically,
is how do we then adopt ourstrategies and think through fitting into the
way they want to talk about thecustomer, or how do we encourage them
to think about how we think aboutthe customer when in fact we should be
just looking at the customer and understandingwhat the customer is actually doing and then
(07:18):
working within that framework. And we'venot really had the opportunity to do that
in any kind of meaningful way todate. And so I think the AI
space we can start with, Okay, these are the collection of people who've
done the thing we're trying to driveforward, whether it's leads or footballer sales
or whatever it might be, andactually explore that group of people and say
(07:38):
what got them here instead of tryingto fit them into these structures that may
be artificial and may not actually applyto their business, and imagine that like
learning actual actually from people's behavior.It's kind of a simple but novel concept.
Well that's you know, it's aninteresting point because I guess this is
really where AI is really powerful,right, because it could it could deal
(08:01):
with the fact that consumers and consumerjourneys may start with a conversation with their
friend and the coffee shop about abrand and or you know, or may
start with them having seen something onAmazon and not necessarily having seen a brand
campaign, etc. Yeah, andAmazon's an interesting one. I was having
(08:22):
this debate with my partner last night. You know, with Amazon, for
example, there's a whole bunch ofthings that you buy when you never even
know what brand it was, youknow, thinking about Sure, you can
buy an object on Amazon where youknow the brand that you've gotten through and
purchased it. But you can alsobuy something now where you're like, oh,
I need a battery pack for mycell phone? Do you know what
brand it is? Or does itcome from Amazon? And a little plastic
baggy with a sticker on it andyou never know. So the awareness familiarity
(08:45):
consideration pieces actually with Amazon, notwith the product itself. And how do
we make sure we account for allof these different nuances and then we we're
thinking about supporting our clients. Soundspretty simple right in practice, there's a
lot of consider rations. Now thatwe have a broader idea of how AI
can help us understand the diverse datapoints that make up the customer journey,
(09:05):
let's talk about what a great AIcentered data strategy might look like. It
starts with data centric AI, aphrase and concept popularized by machine learning pioneer
andrew Ing, it's interesting that wecall it data centric AI. I think
about it is organizing our data forthe AI application. The one example of
(09:35):
how you might think about this aswe spent a lot of time building data
applications that are oriented for retrieval.So if I'm a media manager, I
would like to look at, youknow, how I'm delivering impressions across media
and over time and geography, andso my data engineer goes and builds a
(09:56):
database that's arranged in this way soI can retrieve those answers. Well,
if I actually want to use somepredictive intelligence, that predictive intelligence would like
the data arranged in a way thathelps it recover that consumer journey that we
were talking about earlier. And canthe AI be biased based on what data
(10:16):
it is fat because we know thatthat's an issue when it comes to generating
content or images. Of course,AI can also reduce or eliminate the bias.
So AI is the combination of theinformation that takes in and for generative
AI, the supervision we put aroundthe prediction that was the big the big
(10:37):
sort of invention in things like chatGPT is that the team means transformer,
and this was a supervision advisory frameworkthat took human input to make better predictions.
Now, if that human input isbiased, as humans are biased and
that's where the data comes from,a course, the AI is going to
be biased. But where we knowbiases can exist, we can place you
(11:03):
know, expert judgment on the supervisoryframework to overcome things. So let's take
targeting for example, if we livein a world where we can do a
really good job at targeting people thatare either going to be really responsive to
media, or we're trying to reachlots of people and we end up getting
(11:26):
low value inventory as a result.In the world of measuring the impact of
media, that would generate a biasthat you know, you would either overestimate
the impact of that media because you'regoing after high value or you underestimate the
impact because you're buying you know,remnant inventory, and um, there's ways
(11:50):
to correct for that. So ifyou if you bring that sort of supervisory
framework into your AI, you canovercome the natural biases that exist in data
and measurement. And that's why weneed a bunch of different AIS. There's
not going to be some sort ofgeneral form AI but there needs to be
AI for you know, all thedifferent applications we imagine out there. Well,
(12:11):
in speaking of data, I meaneveryone is talking about CDPs and composable
CDPs and well let's start Kristen maybeanswer for the audience. What a CDP
stand for Customer data platform. It'ssomething our clients are talking about a lot.
In some ways, the conversations thatI've had around it have been people
(12:33):
calling me and saying I want tobuild a CDP and I say, okay,
why and they say, oh,because everyone is and I need one
to me. It feels similar tothe big data conversation we were having a
couple of years ago, which isI need a data warehouse, and lots
of money was spent on a datawarehouse and then it sat and collected dust
and cobwebs and everyone felt like itwas a poor investment. So it's all
(12:56):
the rage right now when it comesto data conversations and what's a composable CDP?
Composable is this idea? You know, CDP is new in the world
of marketing, uh, and manyhave made the investment and built them for
specific use cases, say to helpcall centers. You know, understand what
next best actions are in conversations withcustomers UM or you know, in email
(13:20):
applications. But at the end ofthe day, we're talking about unifying our
view of the customer and the dataabout that UM. So, you know,
different CDPs have become good at differentthings. Either the intake an organization
of data, the ability to segmentcustomers based on that data, using each
for their their their superpowers and composingthem together is the idea behind you know,
(13:46):
composable CDP and to be flexible asyour use cases evolved as you move
up the data maturity curve. Whatwould be your advice to a marketer where
where should they start? They shouldstart internally by talking to their CIO or
are their I tech stakeholders that areoften going to be part of these sort
of initiatives, and open the linesof communications saying well, this is what
(14:09):
we're trying to do as a companywith our customers and with our marketing strategy,
and this is the kind of datathat we think can inform that.
And then you have two people thatare you know, considering all the challenges
around standing something like this up.I think there's a step before that that's
really important. The number of timesI've had conversations with clients where we say,
(14:33):
okay, the goal is we wantto grow sales, and I say,
okay, well what talk to meabout that? What are you thinking?
What are your actual goals where youwanting to go beyond that? And
the clarity on KPI is or evenjust what does success look like for this
company for this client is not somethingthat everybody has in their back pocket.
(14:54):
I think taking a step back andeven asking yourself, you know, what
do you want to do with customersrequires some preliminary work to say here are
some of the challenges, here's thegap in our process. Here's the thing
that isn't working that we need touse communications or data to help fix.
Here are some of the things overhere that are working that we want to
use data to amplify. So Ido think that's where then Michael and I
(15:15):
together work from both a very technicalside and a contextual side to make sure
that we're all aligned on goals,because the CIO isn't going to do that
for you, right The CIO isgoing to help build everything and put it
together, but you need to havea very clear understanding of why and what
and how well, maybe without namingany names, you can give us an
example, So, for example,had a client who had access to online
(15:39):
leads, but not access to leadsto the call center. So if you're
looking at evaluating your media and theonly KPI or output data you have is
online the generation, it's not asurprise that specific channels are going to look
like they're performing better and have abetter ROI because it's connected to an online
activity. So you need to finda way to bring in the call center
(16:02):
data into that conversation because fundamentally thatlead I mean, we're going to in
this case assume it has the samevalue as the online lead, and you're
going to end up skewing your mediaand your marketing plans against your online leads
because of it. And this happensa lot actually, because online e com
data is so much easier to dealwith than some of the offline engagements.
(16:23):
You see. Then collections of mediathings media plans optimize towards your online goals,
but often those online pieces only accountfor a small percentage of the total,
and we get in this horrible cycleof optimizing towards five percent of your
success, which is not going tocreate overall growth. Thanks to AI,
(16:47):
we can now get a much clearerpicture of the data points that make up
the customer journey, and we won'tfall into the trap of overestimating areas with
easier to read data. We're goingto take a quick break and when we
come back, we'll talk about howAI is going to change the way we
work. Welcome back to Partners andPossibility. I'm Brett Marshand. Today I'm
(17:17):
talking with some of our leaders atPlus Company, Kristen Wazniak, SVP of
Customer Success and Michael Cohen, ChiefData and Analytics Officer. We're talking all
about data and AI. We alreadyknow that marketers are finding many uses for
AI and content and creative production,but there's an even bigger opportunity in measurement
management and data. So let's talkabout how this will change the way we
(17:38):
work and how it will impact thepeople in marketing agencies, media agencies and
PR. Here's Kristen. Yeah,So for fun, earlier, I took
all the questions that we thought aboutmaybe tackling today, put them in a
chat GPT just to see what achat GPT and tell me about exactly how
(18:00):
is this going to impact people.What's the impact that AI is going to
have on marketing? And the answersit gave me were very intelligent. Hear
all the different worklow that it couldimpact. What it's not talking about,
though, is the reaction to thosepotential improvements or adjustments. I think the
first thing you're going to see isa lot of discomfort that the folks who
(18:22):
are experienced in the field are goingto say, I would never ever do
that. And it's going to takea moment, I think, for everybody
to eat a piece of humble pieor just take a breath and say,
Okay, maybe I need to trysomething a little bit different. So there's
going to be a lot of discomfortat first, but I do think over
time, what it's going to dois change the way in which we fundamentally
(18:45):
make decisions. We're going to beable to your conversation earlier bread around.
You know, half advertising works,we just don't know which half. There
might be a point where in factthat doesn't matter. It doesn't matter we
don't necessarily know what it does.We just need to know that is working.
So how do we think differently aboutthe structures that we have in place,
(19:06):
the way in which we make decisionsand being okay with a little bit
of risk. Interesting, chat GPTdefinitely did not say any of that.
They said it will automate things.Yeah, you think chat GPT would come
up with a really good answer andwhy it's good for humanity given it's how
important the future of humanity is tochat GPT and its success. But anyways,
(19:26):
I digress. Interested in your pointof view and how AI is going
to change the role that people takein this industry and what we do for
a living and what marketers do fora living, where people, those that
are going to be most successful inthe new world are going to be able
to figure out how to build onthe capability of AI by bringing human judgment
(19:51):
to the table because that's what AIonly gains from humans, and you can
only encode in AI from humans,and that's a that's a growth process.
So as we talk to our teamsand clients and agencies about how to get
(20:11):
the most out of AI through atest and learn framework, it's this process
of test learn and then in codethings that are working in the AI to
then move on to the next higherlevel function of AI and people that can
champion that as their role rather thanI'm going to do the same thing,
(20:33):
but AI is going to do allthe stuff that I don't like to do.
No, AI is going to dothe stuff you like to do as
well. YEA to move on,and those that become good at bringing human
incodable human judgment to AI are reallygoing to catapult in the marketplace as we
rethink the way we organize as firms, our business models, and then the
(20:56):
things that we can offer the marketplace. Well, I know that one of
the powerful things that you guys areworking on, and one are the really
interesting parts about it, is beingable to do scenario planning. Right,
how does that work? You know? I've been on analytics research teams most
of my career and there's a consistenttrend that analytics and research teams are brought
into justify decisions. So this iswhat we did. Now, tell me
(21:19):
why it was the right call,This was the can campaign? Tell me
you know what parts of it weresuccessful? So I think what you're going
to see is analytics become less ofan afterthought and more of a strategic input,
So not using it to justify decisions, but actually using it to inspire
decisions. You know, everybody seemsto be happiest when we get to try
(21:41):
different new things, new opportunities inthe industry, first to market type stuff,
and it's typically a hesitancy to dothat because we can't, you know,
promise that it's going to have areturn suddenly. Now with scenario planning,
with predictions and whatnot, we canstart to say, you know,
this is potentially what it might beable to do for us, and that's
fun. It allows us to experiment, it allows us to try new things,
(22:04):
and it allows the analytics team tocome out from behind, you know,
the curtain and actually be part ofthe main stage. Is that it
just being the set crew? Yeah, you know, I remember that tension
as a client, which is okay, that is a really interesting idea,
but oh my god, that's risky, what you know. And this idea
of being able to test and learnand use data to to to verify that
(22:27):
it's going to work and that it'snot risky, you know, it's not
going to be a waste of money, is super interesting. I think it
can be incredibly empowering, I thinkfor both sides. Right, and then
even if there's a situation where it'struly first to market, and there's not
really a way that we can reliablypredictorur scenario plan around it. Okay,
(22:48):
so we'll do this over here andthat everything else that we're doing, we're
going to make sure in scenario planand predict that we're going to make sure
that this stuff is eighty percent orninety percent of what we're doing over here
will get you your goals. Sorest assure this part over here that we
can absolutely, you know, predictwith a lot of assuredness is going to
get you where you need to go. So we can actually now take these
dollars and try something new and it'sjust greaty right, So there's you know,
(23:11):
there's sort of a sense of commonsecurity on both sides. Let's just
end up talking a little bit aboutsome of the best practices that you guys
are involved in here at plus Company, what we're doing in general in order
to take advantage of this AI revolution. I mean, some of the basic
stuff is about access at your fingertips. And it's interesting to me watching a
(23:34):
collection of folks across the network wantingto use chat, GPT and other systems
to actually help them brainstorm around abrief or a challenge, but doing it
secretly. It's like, oh,I'm not actually allowed to do this.
There's a sort of sense that it'slike cheating still if you ask to ask
chat GPT to provide answers. Soone of the things that we've been doing
(23:56):
is is again a collection of trainingaround critical thinking how to actually interpret us
and augment some of the learnings orbrainstorms that come from chat GPT, but
beyond that, you know, ina more substantial way rethinking measurements. Specifically,
we know that our clients are quitefrankly, we are too, even
(24:17):
in an analytics team, incredibly frustratedby the sheer volume of data that comes
back. And I don't mean justlike data streams, but I mean a
brand lift study from this group,or you know, a proof of concept
from this group, or you know, a campaign summary from this vendor.
Trying to put it all together,and you know, if we're left trying
to connect all the dots and stilldon't feel a hundred percent confident with you
(24:38):
know the question of how the campaigndo, our clients are obviously ten times
more frustrated. So you know wherewe've really started is how do we clean
up that conversation for our clients.How do we really bring it together and
make sure that we can answer withcertainty, with clarity, with transparency.
Here's how your marketing efforts are performingin near real time in a way that
(25:03):
feels understandable and accessible and creates asense of confidence. So that's where we've
really been putting a lot of effort. You can do that on a very
big scale. You can also doit on a very small scale, but
that really is I think one ofthe biggest challenges we need to overcome and
one of the most effective ways toget AI moving from Plus Company right now,
Well, listen, this has beenfascinating, so thank you for joining
(25:26):
me. Thank you for all yourinsights, Good luck on all the work
that you're doing. And I can'twait for you guys both to have John
want to make her role around itas grave and say see somebody solve this
issue. We're on it. Thankyou, guys, Thank you very much,
(25:48):
Thank you for listening to Partners andpossibility. That was my discussion with
two leaders from the Plus Company team, Kristen Wozniak, SVP of Customer Success
and Michael Cohen, Chief Data anAnalytics Officer before we go. One more
thing to consider is around policy andethics when using AI. With a multitude
of platforms and applications we use morethan twenty in plus company, we need
(26:11):
to consider consciously how and when weuse AI and how we protect our own
IP as well as that of ourclients and our creators.
You think chat GPT would come upwith a really good answer and why it's
good for humanity You've given it's howimportant the future of humanity is to chat
GPT and its success. But anyways, I digressed. In a never changing
world, campaigns are all about collaborationand connection, from an initial brief to
(00:24):
the follow through, we explore howinnovative thought leadership makes a campaign success more
than a possibility. Welcome to partnersand Possibility. I'm your host, Brett
Marshaun, CEO of Plus Company.By now, most marketers have dabbled an
AI and introduced aipower tools to theircreative workflows. In our last two episodes,
we discussed how marketers are using thesetools to work smarter and faster.
(00:47):
But one area we haven't talked aboutis data and exciting opportunity for marketers.
Joining me in this episode are twomembers of the Plus Company, t Kristen
Masniak, SVP of Customers six andMichael Cohen, Global Chief Data and Analytics
Officer. We'll go beyond chat GPTand mid journey to examine how data and
generative AI can improve marketing, businessoperations and increase ROI. We'll talk about
(01:15):
how we as marketers can leverage AIto better allocate targeted AD spends. We'll
define a composable CDP and how itand data centric AI will change the way
marketers work. We'll also try toimagine how the future AI will look for
marketers, recognizing it's impossible to predictthe unpredictable world of AI. Michael,
(01:40):
you say something on a regular basis, and it's about the horse versus the
car. Where if you ask peopleto protect and what they want. You
know, in the early nineteen hundredsthey say a faster horse. No one
would have necessarily imagined the car.And so I think there's an element too
of thinking through what it is thatwe want from the future of AI,
and then what will surprise as usgoing forward. Fundamentally, I think it's
(02:02):
important that we keep an eye onour demands, so to speak, or
there'll be a collection of new opportunities, new solutions to meet our demands.
And my demands I don't mean likemy demands, but things that I need
help answering questions, decisions. It'llbe a question of things we don't even
know what they are yet because wecan't quite yet, but hopefully to answer
(02:24):
some of our questions. Yeah,it's I mean, that's an interesting analogy.
I was thinking. I'm reading thisbook about it's called chip Wars,
which I highly recommend. It talksabout sort of the history of the transistor
and and cilicon chips and what it'sdone and Moore's law, et cetera.
And you know, everyone just thoughtit was going to be used in the
(02:46):
beginning for the military, and itwas really when people looked at it beyond
that that they realized that it hadthis incredible power. And it's kind of
the same thing with AI, right, I often say to people, even
though everyone is thinking about generative AIas far as producing text and some people
are thinking about it, you know, around images, not a lot of
(03:07):
people are thinking about it when itcomes to good old fashioned marketing dollars.
And I don't know, for everydollar spent on content, there's a ten
dollars spent on distribution and media.So you know, when you think about
it that way, there's probably tentimes the opportunity when it comes to measurement
and management. How are you seeingit helping with measurement, with management and
with data outputs? You know,the big picture here is that this sort
(03:30):
of prediction. Generative a on whateverwe want to call it, produces information,
whether that's in the form of languageor audio, or image or numbers,
and data answers the questions that wehave allows us to do that in
an iterative way. You might ask, what was my reach for men in
the sixty five to seventy five yearold to age group in southern California,
(03:53):
and how did they respond to aparticular campaign? Like you can layer all
those demensions on it at once.It will synthesize an answer based on the
information it has. It's not requiredto have all the data, and the
reality is that data is imperfect,but prediction can overcome that. One of
(04:15):
the interesting things about what you justsaid, Michael, especially thinking about the
vast collection of clients and regions,is data availability is so mixed. The
generative AIP starts to make it possiblefor clients, for people who haven't been
able to engage in this space before, because they can work with imperfect data
in a more meaningful and reliable way. Right. You know, there's that
(04:38):
famous quote by John Wannamaker, whichis I know half my advertising is working,
I'm just not sure which half,And I think one of the interesting
things about AI and how it canbe used in a predictive way, as
you just both described, is ifpeople can focus on what's working and how
do we do more of that,and unless like is it working or not?
(04:58):
And do I trust it or not? Etc. Tell me a little
bit about how we're using or you'reusing AI now, in particular, not
on the generative side of content,but on this management and measurement side that
we're talking about. Really, thefundamental sort of building block of the of
the marketplace that we aim to influenceis the customer and understanding that entire communication
(05:23):
journey we're having with them. Weknow about all the challenges to filling out
that story, data that exists behindwalled gardens, privacy regulations, you know,
low quality telemetry, just missing pieces. But AI or generative AI can
can solve the puzzle and predict whatthat journey looks like. And now we
(05:46):
have a building block to be ableto um to tell all of these combined
stories about audience, creative media,timing, different KPIs the journey through the
funnel, and do so in away that allows us to build technologies that
answer the questions we have today basedon how mature we are as data and
(06:10):
AI sort of organizations, but alsoanswer the questions of the future that we
don't know because we've built it aroundthe customer, who's always going to be
the building block of the marketplace.I think there's an interesting conversation here around
journeys in the funnel specifically. I'vestruggled with the funnel of my whole career,
and not because you know what isit eighteen ninety eight. It was
(06:32):
developed and stuff has changed since then. But I think more and more as
the industry changes, the way wepurchase things change, the funnel isn't as
straightforward. I don't think we canthrow it away. There's lots of categories
and instances where the funnel and thatlinear journey does make a lot of sense,
and there's other areas now where itreally really doesn't. And if I
work with different clients, everybody comesto the table with their own sort of
(06:56):
linear version of what that journey lookslike. Even if it is a goal,
there's this one model, and theconversation then from the agency side specifically,
is how do we then adopt ourstrategies and think through fitting into the
way they want to talk about thecustomer, or how do we encourage them
to think about how we think aboutthe customer when in fact we should be
just looking at the customer and understandingwhat the customer is actually doing and then
(07:18):
working within that framework. And we'venot really had the opportunity to do that
in any kind of meaningful way todate. And so I think the AI
space we can start with, Okay, these are the collection of people who've
done the thing we're trying to driveforward, whether it's leads or footballer sales
or whatever it might be, andactually explore that group of people and say
(07:38):
what got them here instead of tryingto fit them into these structures that may
be artificial and may not actually applyto their business, and imagine that like
learning actual actually from people's behavior.It's kind of a simple but novel concept.
Well that's you know, it's aninteresting point because I guess this is
really where AI is really powerful,right, because it could it could deal
(08:01):
with the fact that consumers and consumerjourneys may start with a conversation with their
friend and the coffee shop about abrand and or you know, or may
start with them having seen something onAmazon and not necessarily having seen a brand
campaign, etc. Yeah, andAmazon's an interesting one. I was having
(08:22):
this debate with my partner last night. You know, with Amazon, for
example, there's a whole bunch ofthings that you buy when you never even
know what brand it was, youknow, thinking about Sure, you can
buy an object on Amazon where youknow the brand that you've gotten through and
purchased it. But you can alsobuy something now where you're like, oh,
I need a battery pack for mycell phone? Do you know what
brand it is? Or does itcome from Amazon? And a little plastic
baggy with a sticker on it andyou never know. So the awareness familiarity
(08:45):
consideration pieces actually with Amazon, notwith the product itself. And how do
we make sure we account for allof these different nuances and then we we're
thinking about supporting our clients. Soundspretty simple right in practice, there's a
lot of consider rations. Now thatwe have a broader idea of how AI
can help us understand the diverse datapoints that make up the customer journey,
(09:05):
let's talk about what a great AIcentered data strategy might look like. It
starts with data centric AI, aphrase and concept popularized by machine learning pioneer
andrew Ing, it's interesting that wecall it data centric AI. I think
about it is organizing our data forthe AI application. The one example of
(09:35):
how you might think about this aswe spent a lot of time building data
applications that are oriented for retrieval.So if I'm a media manager, I
would like to look at, youknow, how I'm delivering impressions across media
and over time and geography, andso my data engineer goes and builds a
(09:56):
database that's arranged in this way soI can retrieve those answers. Well,
if I actually want to use somepredictive intelligence, that predictive intelligence would like
the data arranged in a way thathelps it recover that consumer journey that we
were talking about earlier. And canthe AI be biased based on what data
(10:16):
it is fat because we know thatthat's an issue when it comes to generating
content or images. Of course,AI can also reduce or eliminate the bias.
So AI is the combination of theinformation that takes in and for generative
AI, the supervision we put aroundthe prediction that was the big the big
(10:37):
sort of invention in things like chatGPT is that the team means transformer,
and this was a supervision advisory frameworkthat took human input to make better predictions.
Now, if that human input isbiased, as humans are biased and
that's where the data comes from,a course, the AI is going to
be biased. But where we knowbiases can exist, we can place you
(11:03):
know, expert judgment on the supervisoryframework to overcome things. So let's take
targeting for example, if we livein a world where we can do a
really good job at targeting people thatare either going to be really responsive to
media, or we're trying to reachlots of people and we end up getting
(11:26):
low value inventory as a result.In the world of measuring the impact of
media, that would generate a biasthat you know, you would either overestimate
the impact of that media because you'regoing after high value or you underestimate the
impact because you're buying you know,remnant inventory, and um, there's ways
(11:50):
to correct for that. So ifyou if you bring that sort of supervisory
framework into your AI, you canovercome the natural biases that exist in data
and measurement. And that's why weneed a bunch of different AIS. There's
not going to be some sort ofgeneral form AI but there needs to be
AI for you know, all thedifferent applications we imagine out there. Well,
(12:11):
in speaking of data, I meaneveryone is talking about CDPs and composable
CDPs and well let's start Kristen maybeanswer for the audience. What a CDP
stand for Customer data platform. It'ssomething our clients are talking about a lot.
In some ways, the conversations thatI've had around it have been people
(12:33):
calling me and saying I want tobuild a CDP and I say, okay,
why and they say, oh,because everyone is and I need one
to me. It feels similar tothe big data conversation we were having a
couple of years ago, which isI need a data warehouse, and lots
of money was spent on a datawarehouse and then it sat and collected dust
and cobwebs and everyone felt like itwas a poor investment. So it's all
(12:56):
the rage right now when it comesto data conversations and what's a composable CDP?
Composable is this idea? You know, CDP is new in the world
of marketing, uh, and manyhave made the investment and built them for
specific use cases, say to helpcall centers. You know, understand what
next best actions are in conversations withcustomers UM or you know, in email
(13:20):
applications. But at the end ofthe day, we're talking about unifying our
view of the customer and the dataabout that UM. So, you know,
different CDPs have become good at differentthings. Either the intake an organization
of data, the ability to segmentcustomers based on that data, using each
for their their their superpowers and composingthem together is the idea behind you know,
(13:46):
composable CDP and to be flexible asyour use cases evolved as you move
up the data maturity curve. Whatwould be your advice to a marketer where
where should they start? They shouldstart internally by talking to their CIO or
are their I tech stakeholders that areoften going to be part of these sort
of initiatives, and open the linesof communications saying well, this is what
(14:09):
we're trying to do as a companywith our customers and with our marketing strategy,
and this is the kind of datathat we think can inform that.
And then you have two people thatare you know, considering all the challenges
around standing something like this up.I think there's a step before that that's
really important. The number of timesI've had conversations with clients where we say,
(14:33):
okay, the goal is we wantto grow sales, and I say,
okay, well what talk to meabout that? What are you thinking?
What are your actual goals where youwanting to go beyond that? And
the clarity on KPI is or evenjust what does success look like for this
company for this client is not somethingthat everybody has in their back pocket.
(14:54):
I think taking a step back andeven asking yourself, you know, what
do you want to do with customersrequires some preliminary work to say here are
some of the challenges, here's thegap in our process. Here's the thing
that isn't working that we need touse communications or data to help fix.
Here are some of the things overhere that are working that we want to
use data to amplify. So Ido think that's where then Michael and I
(15:15):
together work from both a very technicalside and a contextual side to make sure
that we're all aligned on goals,because the CIO isn't going to do that
for you, right The CIO isgoing to help build everything and put it
together, but you need to havea very clear understanding of why and what
and how well, maybe without namingany names, you can give us an
example, So, for example,had a client who had access to online
(15:39):
leads, but not access to leadsto the call center. So if you're
looking at evaluating your media and theonly KPI or output data you have is
online the generation, it's not asurprise that specific channels are going to look
like they're performing better and have abetter ROI because it's connected to an online
activity. So you need to finda way to bring in the call center
(16:02):
data into that conversation because fundamentally thatlead I mean, we're going to in
this case assume it has the samevalue as the online lead, and you're
going to end up skewing your mediaand your marketing plans against your online leads
because of it. And this happensa lot actually, because online e com
data is so much easier to dealwith than some of the offline engagements.
(16:23):
You see. Then collections of mediathings media plans optimize towards your online goals,
but often those online pieces only accountfor a small percentage of the total,
and we get in this horrible cycleof optimizing towards five percent of your
success, which is not going tocreate overall growth. Thanks to AI,
(16:47):
we can now get a much clearerpicture of the data points that make up
the customer journey, and we won'tfall into the trap of overestimating areas with
easier to read data. We're goingto take a quick break and when we
come back, we'll talk about howAI is going to change the way we
work. Welcome back to Partners andPossibility. I'm Brett Marshand. Today I'm
(17:17):
talking with some of our leaders atPlus Company, Kristen Wazniak, SVP of
Customer Success and Michael Cohen, ChiefData and Analytics Officer. We're talking all
about data and AI. We alreadyknow that marketers are finding many uses for
AI and content and creative production,but there's an even bigger opportunity in measurement
management and data. So let's talkabout how this will change the way we
(17:38):
work and how it will impact thepeople in marketing agencies, media agencies and
PR. Here's Kristen. Yeah,So for fun, earlier, I took
all the questions that we thought aboutmaybe tackling today, put them in a
chat GPT just to see what achat GPT and tell me about exactly how
(18:00):
is this going to impact people.What's the impact that AI is going to
have on marketing? And the answersit gave me were very intelligent. Hear
all the different worklow that it couldimpact. What it's not talking about,
though, is the reaction to thosepotential improvements or adjustments. I think the
first thing you're going to see isa lot of discomfort that the folks who
(18:22):
are experienced in the field are goingto say, I would never ever do
that. And it's going to takea moment, I think, for everybody
to eat a piece of humble pieor just take a breath and say,
Okay, maybe I need to trysomething a little bit different. So there's
going to be a lot of discomfortat first, but I do think over
time, what it's going to dois change the way in which we fundamentally
(18:45):
make decisions. We're going to beable to your conversation earlier bread around.
You know, half advertising works,we just don't know which half. There
might be a point where in factthat doesn't matter. It doesn't matter we
don't necessarily know what it does.We just need to know that is working.
So how do we think differently aboutthe structures that we have in place,
(19:06):
the way in which we make decisionsand being okay with a little bit
of risk. Interesting, chat GPTdefinitely did not say any of that.
They said it will automate things.Yeah, you think chat GPT would come
up with a really good answer andwhy it's good for humanity given it's how
important the future of humanity is tochat GPT and its success. But anyways,
(19:26):
I digress. Interested in your pointof view and how AI is going
to change the role that people takein this industry and what we do for
a living and what marketers do fora living, where people, those that
are going to be most successful inthe new world are going to be able
to figure out how to build onthe capability of AI by bringing human judgment
(19:51):
to the table because that's what AIonly gains from humans, and you can
only encode in AI from humans,and that's a that's a growth process.
So as we talk to our teamsand clients and agencies about how to get
(20:11):
the most out of AI through atest and learn framework, it's this process
of test learn and then in codethings that are working in the AI to
then move on to the next higherlevel function of AI and people that can
champion that as their role rather thanI'm going to do the same thing,
(20:33):
but AI is going to do allthe stuff that I don't like to do.
No, AI is going to dothe stuff you like to do as
well. YEA to move on,and those that become good at bringing human
incodable human judgment to AI are reallygoing to catapult in the marketplace as we
rethink the way we organize as firms, our business models, and then the
(20:56):
things that we can offer the marketplace. Well, I know that one of
the powerful things that you guys areworking on, and one are the really
interesting parts about it, is beingable to do scenario planning. Right,
how does that work? You know? I've been on analytics research teams most
of my career and there's a consistenttrend that analytics and research teams are brought
into justify decisions. So this iswhat we did. Now, tell me
(21:19):
why it was the right call,This was the can campaign? Tell me
you know what parts of it weresuccessful? So I think what you're going
to see is analytics become less ofan afterthought and more of a strategic input,
So not using it to justify decisions, but actually using it to inspire
decisions. You know, everybody seemsto be happiest when we get to try
(21:41):
different new things, new opportunities inthe industry, first to market type stuff,
and it's typically a hesitancy to dothat because we can't, you know,
promise that it's going to have areturn suddenly. Now with scenario planning,
with predictions and whatnot, we canstart to say, you know,
this is potentially what it might beable to do for us, and that's
fun. It allows us to experiment, it allows us to try new things,
(22:04):
and it allows the analytics team tocome out from behind, you know,
the curtain and actually be part ofthe main stage. Is that it
just being the set crew? Yeah, you know, I remember that tension
as a client, which is okay, that is a really interesting idea,
but oh my god, that's risky, what you know. And this idea
of being able to test and learnand use data to to to verify that
(22:27):
it's going to work and that it'snot risky, you know, it's not
going to be a waste of money, is super interesting. I think it
can be incredibly empowering, I thinkfor both sides. Right, and then
even if there's a situation where it'struly first to market, and there's not
really a way that we can reliablypredictorur scenario plan around it. Okay,
(22:48):
so we'll do this over here andthat everything else that we're doing, we're
going to make sure in scenario planand predict that we're going to make sure
that this stuff is eighty percent orninety percent of what we're doing over here
will get you your goals. Sorest assure this part over here that we
can absolutely, you know, predictwith a lot of assuredness is going to
get you where you need to go. So we can actually now take these
dollars and try something new and it'sjust greaty right, So there's you know,
(23:11):
there's sort of a sense of commonsecurity on both sides. Let's just
end up talking a little bit aboutsome of the best practices that you guys
are involved in here at plus Company, what we're doing in general in order
to take advantage of this AI revolution. I mean, some of the basic
stuff is about access at your fingertips. And it's interesting to me watching a
(23:34):
collection of folks across the network wantingto use chat, GPT and other systems
to actually help them brainstorm around abrief or a challenge, but doing it
secretly. It's like, oh,I'm not actually allowed to do this.
There's a sort of sense that it'slike cheating still if you ask to ask
chat GPT to provide answers. Soone of the things that we've been doing
(23:56):
is is again a collection of trainingaround critical thinking how to actually interpret us
and augment some of the learnings orbrainstorms that come from chat GPT, but
beyond that, you know, ina more substantial way rethinking measurements. Specifically,
we know that our clients are quitefrankly, we are too, even
(24:17):
in an analytics team, incredibly frustratedby the sheer volume of data that comes
back. And I don't mean justlike data streams, but I mean a
brand lift study from this group,or you know, a proof of concept
from this group, or you know, a campaign summary from this vendor.
Trying to put it all together,and you know, if we're left trying
to connect all the dots and stilldon't feel a hundred percent confident with you
(24:38):
know the question of how the campaigndo, our clients are obviously ten times
more frustrated. So you know wherewe've really started is how do we clean
up that conversation for our clients.How do we really bring it together and
make sure that we can answer withcertainty, with clarity, with transparency.
Here's how your marketing efforts are performingin near real time in a way that
(25:03):
feels understandable and accessible and creates asense of confidence. So that's where we've
really been putting a lot of effort. You can do that on a very
big scale. You can also doit on a very small scale, but
that really is I think one ofthe biggest challenges we need to overcome and
one of the most effective ways toget AI moving from Plus Company right now,
Well, listen, this has beenfascinating, so thank you for joining
(25:26):
me. Thank you for all yourinsights, Good luck on all the work
that you're doing. And I can'twait for you guys both to have John
want to make her role around itas grave and say see somebody solve this
issue. We're on it. Thankyou, guys, Thank you very much,
(25:48):
Thank you for listening to Partners andpossibility. That was my discussion with
two leaders from the Plus Company team, Kristen Wozniak, SVP of Customer Success
and Michael Cohen, Chief Data anAnalytics Officer before we go. One more
thing to consider is around policy andethics when using AI. With a multitude
of platforms and applications we use morethan twenty in plus company, we need
(26:11):
to consider consciously how and when weuse AI and how we protect our own
IP as well as that of ourclients and our creators.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!