Latest / The Marketing Meeting with Itir Eraslan / 41. Marketing with the "House of AI" with Anindya Ghose
Transcript
- 0:00Anindya Ghose: Today, at least, if not five years ago, if you're not
- 0:03doing performance marketing, you are making a big strategic mistake.
- 0:07My, my first advice to companies who haven't really embraced
- 0:11performance marketing or digital marketing is there is no valid
- 0:14reason today not to be in that space.
- 0:17But if you don't, like you rightly said, it is a disadvantage.
- 0:20You will be left behind.
- 0:22Make no mistake.
- 0:23Today's world, it's not a choice.
- 0:25Itir Eraslan: Hi, this is the marketing meeting, and I'm your host Itir Eraslan.
- 0:30Every two weeks I meet with experts and we talk about topics related to
- 0:34brands, marketing, and businesses.
- 0:36We sometimes add random lifestyle topics too.
- 0:39I hope you enjoy the show.
- 0:45Welcome to the Marketing Meeting Podcast.
- 0:47And I'm happy to host Anindya Ghose today.
- 0:50He's a professor at NYU Stern, specializing in quantitative
- 0:54marketing and tech economics.
- 0:56His first book, TAP, Unlocking the Mobile Economy, is one of my
- 1:00favorite books, but it's also a bestseller and an award winning book.
- 1:05Uh, now I'm really excited that, uh, his second book is coming up,
- 1:09uh, which is tribe, maximizing well being in the age of AI.
- 1:13And we'll talk about this, this episode.
- 1:15Welcome, Anindya.
- 1:16Anindya Ghose: Thank you.
- 1:17Thank you for having me.
- 1:17Looking forward to the conversation.
- 1:20Itir Eraslan: Um, I learned that you are an avid high altitude mountaineer.
- 1:25Uh, what are some of the places that you've climbed?
- 1:29Anindya Ghose: Yeah, that's true.
- 1:30I'm very, very, you know, I guess, passionate about mountaineering.
- 1:33Um, I have climbed in all five continents.
- 1:35Uh, you know, the Himalayas, Andes, Alps, uh, lots in the Rockies and Cascades.
- 1:40And of course there's Kilimanjaro.
- 1:42Um, yeah, I mean, I guess, you know, I started climbing when I was 22,
- 1:46uh, my undergrad education was in the Northern part of India, which
- 1:50is very close to the Himalayas.
- 1:52And so basically every other weekend I disappear in the mountains and
- 1:56I realized that there is a certain draw, a certain appeal to the mountain
- 2:00that really, You know, his song.
- 2:01And I think it was just, uh, I guess natural that the
- 2:05love, the love was mutual.
- 2:07Itir Eraslan: And you said you climbed with your, uh, daughter lately.
- 2:12Anindya Ghose: Yeah.
- 2:12So my daughter, uh, she's 15 when she was four years old, we took
- 2:16out of the Alps in Europe and.
- 2:19You know, she did a a moderate climb pretty effortlessly and I realized
- 2:23both my wife and I realized like maybe she's going to be A natural too.
- 2:27So ever since she's been four, she's been doing a lot of mostly
- 2:30like long day hikes and multi day tracks But I think Last year, she
- 2:34climbed Mount Fuji in Japan with me.
- 2:37And then, uh, she climbed, uh, 14 hour year in Colorado with me.
- 2:40Uh, and then this year, we climbed Tibet.
- 2:43And she, you know, reached her personal best, almost, uh, 14, 500 feet.
- 2:48Itir Eraslan: Wow.
- 2:49Uh, it's, it's exciting to hear that and then maybe later on after the
- 2:53episode, I'm going to ask you how can we go into climbing in New York City?
- 3:00I, I'm just looking all these, uh, glass cleaners nowadays,
- 3:02you know, it's summertime and everyone is cleaning their doors.
- 3:06And, uh, when sometimes when I see, see them sitting on just like one plate.
- 3:10It's just like scaring me a lot.
- 3:12It's
- 3:12Anindya Ghose: a tough job.
- 3:13They have a tough, risky job.
- 3:15You know, I, I appreciate the work they do and they have to
- 3:18be harnessed up all the time.
- 3:20And it's a really risky and tough job.
- 3:22So a lot of kudos to those people.
- 3:25Itir Eraslan: Yeah.
- 3:25Yeah, I agree.
- 3:27So from there, I'm jumping on to the topic, which is AI.
- 3:31Uh, and before I, uh, you know, ask questions about the book, upcoming
- 3:35book, can you tell me how does.
- 3:37AI differ from previous technologies, from its technological advancements?
- 3:42And how is it similar?
- 3:45Anindya Ghose: Sure.
- 3:45So, um, I think of AI as another GPT, a general purpose technology.
- 3:51Uh, you know, every new GPT comes with its own set of challenges and opportunities.
- 3:56You know, by construction, they often like redefine industries.
- 4:00They make new farms.
- 4:01They make certain jobs obsolete, but they also increase productivity.
- 4:05What is different about AI is that it's significantly more layered and
- 4:09more complex and intangible than prior GPTs, you know, like think
- 4:14about electricity or cars, right?
- 4:17Uh, when people talk about electricity, you could see the light switching
- 4:20on and off, but you can't see AI.
- 4:23You know, when people talk about transportation and computing, you can
- 4:26see a computer working, you can see the cars moving, you don't see AI.
- 4:29So AI is more intangible.
- 4:32And so this is the reason why we see these extreme narratives that swing
- 4:37from dystopia like AI is going to take over all our jobs to utopia, you
- 4:43know, depending on who's narrating.
- 4:44And so, um, you know, that's sort of the reason why we wrote this book to kind of
- 4:48lift the hood of AI for regular people.
- 4:51I'm using the context of how it's actually embedded in, you know, everyday life.
- 4:56Um, in things to care about like health and education and work and relationships.
- 5:00I can talk more about
- 5:03Itir Eraslan: that.
- 5:03What was the driving force to write this book thrive maximizing
- 5:08well being in the age of AI?
- 5:09Because I was curious about the.
- 5:12Especially the second part, maximizing well being,
- 5:15Anindya Ghose: uh,
- 5:15Itir Eraslan: what made you put that headline to the book's name?
- 5:20Anindya Ghose: Right.
- 5:20So, so the, the book is a joint work with me and professor Ravi Bhatna.
- 5:25And both Ravi and I have been immersed in the AI space for the last 20 years,
- 5:29you know, through consulting and research and other kinds of projects.
- 5:33So we've had a very hands on experience and actually coding,
- 5:37building the models, the classifiers.
- 5:39So we've seen the good, the bad and the ugly.
- 5:41And we realize that all around the world, when people talk about AI,
- 5:45the only narrative we see is this.
- 5:48Negative fear mongering is kind of a cottage industry that's trying to
- 5:52create a lot of fear And so we're not dismissing that they can be negative,
- 5:56but there's a lot of positives and so what both of you and I realized there
- 6:01is a market there should be a demand for a book that also Leaving the media's
- 6:06hunt for eyeballs and clicks, you know, we know that negativity sells, right?
- 6:10So that's why it's easy to write negative stories But we also feel that look ai
- 6:14is here to stay and we believe that an informed and educated Citizenry is
- 6:19the missing ingredient for us to get ai to work for the benefit of society,
- 6:24you know
- 6:25Anindya Ghose: For instance people don't realize but we actually give
- 6:27a lot of details in our book that ai is actually keeping us more safe
- 6:32Then what many people would imagine.
- 6:34So that's sort of the motivation for writing this book.
- 6:38Itir Eraslan: In the book, uh, you talk a lot about the, um,
- 6:42House of AI, uh, framework.
- 6:44Right.
- 6:44Itir Eraslan: Uh, and I read an article this morning about that, uh, online.
- 6:49It was like an introduction to House of AI.
- 6:51And I would like to touch base on House of AI.
- 6:53What that is.
- 6:54Uh, that mean and how it works.
- 6:56Anindya Ghose: Sure.
- 6:57Sure.
- 6:57So, so that is one of my favorite parts of the book.
- 7:00Uh, and partly because it comes Mai's joint hands on work with companies.
- 7:05So.
- 7:06You know, one prevalent misconception about AI is that people equate
- 7:11AI with something like ChatGPT.
- 7:15So, ChatGPT is an indeed an impressive large language model.
- 7:19It's developed by OpenAI and Microsoft and so on.
- 7:21But that's just one small facet of artificial intelligence.
- 7:25I've worked with more than 50 companies at this point and in
- 7:29different parts of the world.
- 7:30In our experience, AI encompasses a much wider range of, you know, uh,
- 7:35technologies and applications, including like Additional machine learning and like
- 7:39involving prediction and optimization, causal analytics, prescriptive
- 7:43optimization, uh, deep learning.
- 7:46So what we did is we said, Hey, look, uh, I also run the, uh, AI program
- 7:51at NYU, which is part of my day job.
- 7:53Uh, being a professor, I am the director of the master's AI program.
- 7:57And so in that capacity, also, I have to be heavily hands on involved
- 8:01in the real world with companies.
- 8:03Because our students do these capstone projects that are year long projects.
- 8:07Uh, and these are not just like some, uh, artificial data and analytics project.
- 8:12These are real companies where they solve real problems.
- 8:15And so in doing all of that, what we realize is that we
- 8:20need a proper framework to help organizations adopt and adapt to AI.
- 8:28And so what the house of AI does is that it, it provides a strong
- 8:32foundation, uh, at the bottom, the foundational layer is data engineering.
- 8:37Um, and you know, between him and I, like we have like two, Four decades of
- 8:41experience and we believe both of us believe that between 60 to 70 Of the
- 8:46project time in analytics should be in data engineering Okay, so if you're
- 8:52wondering what is data engineering data engineering means cleaning the data
- 8:56aggregating the data You know curating it in a way where you account for
- 9:01missing variables missing observations, you know, because If you don't do the
- 9:05cleaning of the data, then there's garbage in garbage out problem in AI, right?
- 9:09To be able to make meaningful inferences from your AI models, you need to be able
- 9:12to clean the data And so we strongly believe that two thirds of your time
- 9:16should be spent in that And then the rest of the house of AI is basically
- 9:23placing the spotlight on translation very cogent translation of AI applications
- 9:30into business and societal value in a fair and an equitable way Uh,
- 9:36and so there are these four pillars of AI, which I can talk about it in
- 9:39Itir Eraslan: depth.
- 9:41Just one thing, Ben, you say two thirds of this should be spent in data
- 9:46engineering, which is like cleaning the data, harnessing the data.
- 9:49Does that mean like when a company adopts AI, uh, that house of AI work
- 9:55should be mainly about two thirds of it, mainly about data cleaning?
- 9:59Or Does it involve, let's say if marketing is using AI, then it means like two thirds
- 10:04of marketing needs to be data engineering.
- 10:07Right.
- 10:08Can you elaborate on that?
- 10:09Yeah, it's
- 10:09Anindya Ghose: a very, very good question.
- 10:10Actually, the answer is both.
- 10:12So it depends on, you know, the structure of the organization and how
- 10:16integrated the different disciplines are versus many organizations have
- 10:22marketing and accounting and operations and advertising and finance as silos.
- 10:27And so they have they do not have a centralized data infrastructure
- 10:30They have siloed infrastructure on other and many organizations have
- 10:34an integrated centralized data base or data infrastructure where All
- 10:39disciplines pull from that infrastructure
- 10:42So
- 10:43Anindya Ghose: if you are the farmer, that means you will have to
- 10:45do the data cleaning job for each individual functional discipline
- 10:49Okay, so marketing will have to do it on its own In operation,
- 10:53you'll have to do it on its own.
- 10:54Finance, you'll have to do it on its own.
- 10:56But if you are centralized, right, then you have to do it in one place and then
- 11:00every discipline can pull the clean data and then they can do the model
- 11:05building, you know, the prescriptive analytics, the causal analytics.
- 11:09the descriptive, all of that can be done after that.
- 11:11So in other words, what you're saying is irrespective of whether you are
- 11:15centralized or siloed, okay, you have to spend two thirds of your time, in
- 11:21our opinion, two thirds of your time doing the data engineering, cleaning
- 11:24the data, curating it, you know, creating it, et cetera, making it.
- 11:28Processable so that these AI based models can then plug and play.
- 11:34Itir Eraslan: In terms of, when you say about data engineering,
- 11:36does it, uh, does it vary?
- 11:38Because I'm asking, uh, sometimes 101 AI questions.
- 11:42So bear with me on that.
- 11:44But, uh, Does it differ from one company to the other on, for example, let's
- 11:51say two companies, they are doing both centralized data engineering, but, uh,
- 11:56the way that how you data engineer your data, does it depend on your, you know,
- 12:00strategy on your goal on what you want to get out of this data or that data
- 12:05engineering and cleansing the data is.
- 12:08Is that very, uh, like plug and play work?
- 12:11Anindya Ghose: No, it has to be customized.
- 12:13So it has to be customized.
- 12:15Let's just take marketing as an example.
- 12:17You know, B2C marketing, as you very well know, is very
- 12:19different from B2B marketing.
- 12:21And part of the reason is because if you're doing data driven marketing, you
- 12:25have to start with a unit of analysis.
- 12:28In b2b the unit of analysis is typically another decision maker, which
- 12:32is a company not a consumer in b2c The unit of analysis is typically a
- 12:37customer level a consumer level data.
- 12:40So just based on that itself There's going to be meaningful differences in
- 12:44how you do data engineering because you know When you have disaggregate data
- 12:50and b2c level you can do a lot more with disaggregate data as opposed to
- 12:54aggregated data in b2b In other words You cannot disaggregate aggregate data, but
- 13:00you can aggregate disaggregate data, and I don't know if this is clear, but you
- 13:03can basically go from a smaller unit to a more broader unit, but you cannot go
- 13:08from a broader unit to a smaller unit.
- 13:10You know, people often ask me like, why is B2B marketing behind B2C?
- 13:15That is the reason.
- 13:16The reason is because of the fundamental infrastructure of the unit of data.
- 13:21Okay?
- 13:22Mm-Hmm.
- 13:22. So B2C has a lot more, you know, uh, juice or fuel because they are
- 13:27individual customer level data.
- 13:29You can do a lot more with that.
- 13:31Itir Eraslan: Okay.
- 13:31Yeah.
- 13:31That, that makes sense right now, um, my links question is one, after the
- 13:36data engineering, I know that there's like the descriptive, predictive,
- 13:40casual and prescriptive stage, right?
- 13:43Do, do you call them stage or is it like.
- 13:46Segments under the house.
- 13:48Anindya Ghose: Good question.
- 13:48I don't think of them as stages.
- 13:50Mostly, I think of them as individual pillars.
- 13:54That being said, there is some sequence.
- 13:57So now in my recommendation in the in the book, also, we say the
- 14:00first pillar should be descriptive.
- 14:02At this point, you're not building any fancy models.
- 14:05You're essentially doing what we call ice fishing, you know, fishing in
- 14:08a frozen lake looking for one fish.
- 14:11Meaning that you're exploring the data, look for some patterns, thinking about
- 14:14what questions can you answer that?
- 14:16That's a descriptive pillar.
- 14:18Then the next one is either predictive or causal.
- 14:21Predictive means, uh, what will happen next?
- 14:24Okay, so descriptive is answering one simple question, what has happened so far?
- 14:29And predictive will answer the question, what will happen next?
- 14:32And
- 14:33Anindya Ghose: causal will answer the question.
- 14:34Why did something happen?
- 14:36And then finally is prescriptive, which is what should we do next?
- 14:41But that's the sequence we go.
- 14:42What has happened so far?
- 14:43What can happen next?
- 14:45Why did it happen?
- 14:47And then what should we do?
- 14:48Okay, so those are four And
- 14:52Itir Eraslan: I mean, thinking about the explanations that you gave me, I feel
- 14:56that especially on the prescriptive side, then human beings will play an important
- 15:01role because then how should we respond is something that's probably AI would
- 15:06be able to suggest, but then it's up to the human being to decide based on.
- 15:12Strategic level, right?
- 15:14Anindya Ghose: Yeah, so that's an interesting question.
- 15:15You know, uh, one of my co authors, uh, professor Avi Goldbart he wrote
- 15:19a very interesting book with a lot of people where they said basically
- 15:24that as the prediction power of this economy improves The value of intuition
- 15:31actually goes up not goes down And so they complementarity between prediction
- 15:37and human intuition and I also think the same they're not substitutes You
- 15:41Uh, I do think that as AI becomes more sophisticated, there are going to be
- 15:46many elements and many contexts in which human intervention becomes more
- 15:50important, not less important, which is the same reason why we believe that
- 15:54we had to write this book to dispel the fear that AI will just replace jobs.
- 16:00They're not going to replace jobs.
- 16:02They will obviously, some jobs become obsolete, but they also create new jobs.
- 16:07In addition, the existing jobs.
- 16:09They will simply change the role of the individual human being.
- 16:12You know, a good example I give is ATM machines.
- 16:16You know, 20 years ago, when ATM machines came everybody around us
- 16:19told, Oh, you know, the tellers, the cash tellers are gone.
- 16:22Their jobs are gone.
- 16:23They're
- 16:23Anindya Ghose: not gone.
- 16:24Their jobs are not different.
- 16:26So that's how I think about AI and humans.
- 16:29Itir Eraslan: Um, that was an, that was an interesting example from yesterday.
- 16:33I was.
- 16:33You know, questioning the meaning of life, uh, while I was waiting
- 16:37for a live chat to happen.
- 16:39As a marketer, I'm familiar with AI and so on, but whenever I open a chatbot,
- 16:44which is a part of an AI outcome, it's like a product at the end of the day.
- 16:48But when I open a chatbot, uh, I was trying to change a flight, uh, in United.
- 16:53Uh, and then I, I directly always write, connects me to live chats and I write it
- 16:59five times if I, it doesn't connect to me.
- 17:02I just don't want to deal with chatbots.
- 17:04And I said, I mean, in that chatbot world, there should be some more evolution.
- 17:08Yes.
- 17:10Anindya Ghose: And maybe United is not the best airline.
- 17:14You know that anyway, right?
- 17:15So I think part of the problem is that we are dealing with an
- 17:19organization that's not necessarily the most sophisticated when it comes
- 17:22to adopting technologies and AI, but
- 17:24your,
- 17:24Anindya Ghose: your, your Roger point is very well taken, which is that I think
- 17:27we still have a lot of room to improve.
- 17:30Our interaction with these bots and um, I have seen some some organizations go
- 17:36a long way especially in the far east in like So like India and Singapore,
- 17:41uh, they are much more sophisticated than actually back in the West.
- 17:45Itir Eraslan: And I think in that case, like the adoption of AI and the trust
- 17:49around AI is an important one, because even if the chatbot was offering me
- 17:54the best, and I never tried, I mean, like, I never tried using chatbot
- 17:57for United to see if it can help me with the refund and the change of
- 18:01the flight, uh, because I don't trust it yet, uh, it's not about that.
- 18:06I know United States.
- 18:07chatbot is working or not.
- 18:09It's not about that.
- 18:10It's my general perception of chatbots.
- 18:13Uh, and I think in that sense, Even if the technology advances and if
- 18:17it's the best chatbot, then people's perception in using the technology will
- 18:22have to change, which will probably take a little bit more time and for
- 18:26the technology to be adapted as well.
- 18:29Um, you mentioned on top of the house, these, there's AI powered
- 18:33society and underneath is the AI.
- 18:35ethical, equitable, explainable, and fair AI.
- 18:38Right.
- 18:39Itir Eraslan: So this is the part that a lot of people are discussing right
- 18:42now because, you know, we can already imagine the bad use cases of AI.
- 18:50And then people are worried about that a lot.
- 18:53But for you, an equitable, fair AI is how do you think about that?
- 18:58Anindya Ghose: Yeah.
- 18:58So technically speaking, the solution is not that difficult solution is to minimize
- 19:03or remove algorithmic biases, right?
- 19:05So what does that mean?
- 19:06What that means?
- 19:06You see, an algorithm is basically as good or as bad as the data
- 19:12set on which it is trained.
- 19:13Okay, so a good data scientist before they train the algorithm, they
- 19:18will actually explore a lot of time in the data looking for outliers,
- 19:22outliers, both on the left tail and the right tail of the distribution.
- 19:25Okay.
- 19:26Because these biases come from the outliers the biases don't come from
- 19:29the mean or the averages, right?
- 19:31They come from several standard deviations on the right or the left And so a good
- 19:34data scientist will basically explore and remove these outliers or at least try to
- 19:38understand the reason for the outliers So the technical solution to algorithmic
- 19:43bias is really straightforward.
- 19:45It's not as complicated as you know People think it is and so that's what I
- 19:49would like when I consult for companies uh in and in this issue I start with the
- 19:53ground level and say Hey Uh, let's sit down with your data science team and spend
- 19:58a lot of time exploring these outliers in the data because I want to make sure
- 20:04that we eliminate or mitigate the biases right off the top of the pack Then we
- 20:09are on the right path towards a fair and equitable and more ethical AI, right?
- 20:14Because these outliers have been removed.
- 20:16So that's the very first step Beyond that, then, of course, as the new data
- 20:20comes in, it's not a one shot process.
- 20:23You have to continually explore the new data and make sure that you continue
- 20:28to mitigate these algorithmic biases in the new data that's generated as well.
- 20:32So it's an ongoing process.
- 20:34Itir Eraslan: And I think in that sense, too.
- 20:35There is a lot of work to do for the data engineers and also from the management
- 20:40up to the data engineers to see, okay, what's for our company, what's equitable,
- 20:44what's fair, and what are the filters that the data team should be following as
- 20:51a company, not to fall into that traps.
- 20:54So, I mean, as a marketer, I have a question.
- 20:57So we have the house of AI framework and I'm a marketer.
- 21:01How can I apply this house of AI framework to My function yeah,
- 21:08Anindya Ghose: so, uh, you know, i'm also a professor of marketing and I
- 21:11do a lot of marketing consulting So we have applied the house of ai in
- 21:15like dozens of projects at this point.
- 21:18I'll start with my favorite one My favorite one is marketing mixed
- 21:22modeling followed by multi touch attribution modeling Those are my two
- 21:27favorite ones and they're favorite because they're heavily quantitative
- 21:31You know, they have elements of both causal inference predictive modeling
- 21:36As well as prescriptive modeling and with with one caveat which i'll come
- 21:40to you So it covers descriptive because you first have to explore what are the
- 21:44inputs to your marketing campaigns, you know, then you look at What might
- 21:49happen next you look at substitution?
- 21:51So like when I work with companies I am taking their google campaign data
- 21:55facebook campaign data twitter tiktok linkedin pinterest all of the data And
- 22:00i'm looking for substitution possibilities to see how do I increase the roi?
- 22:03You Of this particular company's campaign.
- 22:06Then I design incrementality tests, you know, for causal inference.
- 22:10I do these randomized control experiments, AB testing.
- 22:14These are called instrumentality tests, which also helps me with this question of
- 22:17how do I maximize the ROI for my client?
- 22:20By substituting in different ad campaigns between facebook and google or google
- 22:26and tiktok or tiktok and twitter, you know And so the one caveat I said where
- 22:30you can also do prescriptive analytics for marketing is In budget allocation.
- 22:35So one of the most complex problems for marketer is how do I allocate?
- 22:39The budget right across these different channels.
- 22:42I have TV and billboard and internet and mobile and, you
- 22:46know, direct newspapers, right?
- 22:48So how do I look at the budget?
- 22:49So I've done that kind of work maybe 15, 20 times now for different companies.
- 22:55And so the prescriptive analytics provides the solution to budget allocation
- 23:00because prescriptive analytics, the foundation of toolkit is optimization.
- 23:04So the only way you can do a good job with budget allocation is if you
- 23:08can optimize the budget condition on these constraints So that's
- 23:12just like one or two good examples.
- 23:14We've done this with attribution Customers both from meta to youtube to
- 23:19instagram linkedin right every session.
- 23:22They open a new touch point, right?
- 23:24Itir Eraslan: Yeah, so
- 23:24Anindya Ghose: the consumer path to purchase journey is no
- 23:26longer linear or sequential.
- 23:28It's very very iterative.
- 23:30It is non linear It is non sequential people go back and forth, you know You
- 23:33That marketing funnel, we used to talk about awareness, interest, desire, action.
- 23:38It's not like that anymore.
- 23:39It's like people skip stages, you know, people, uh, go from awareness
- 23:43to action and then you go back to interest and they go from interest to
- 23:46desire, then they go back to awareness so that old marketing funnel, right,
- 23:51does not exist the way it is, it's very nonlinear, it's non sequential.
- 23:55People skip states to be able to isolate how should this consumer
- 24:00be targeted at each stage.
- 24:02We are now using these AI based models, uh, using MTA and MMM
- 24:08Itir Eraslan: and so on.
- 24:09But it means like, uh, the data that you feed in, since you're going to do quite
- 24:14quantitative at the beginning, the data should be, should be measurable, which
- 24:18means like, I'm just thinking if for a company that has been doing performance
- 24:22marketing, advertising, you know, giving ads to social media and a programmatic,
- 24:27they would have a lot of data.
- 24:29So it gives them an unfair advantage for the companies that haven't done
- 24:34so much of performance marketing and that they are doing like, you know,
- 24:38influencer marketing, uh, Well, some of it can be measured though But like
- 24:43events and so on I would assume that it's especially Applicable for the performance
- 24:48marketing and online advertising, right?
- 24:51Anindya Ghose: Yeah, I mean, you know, I actually think that Today at least uh,
- 24:54if not five years ago if you're not doing performance marketing You are making a
- 24:59big strategic mistake So that's sort of the lens where I work on that my first
- 25:04would have advised to companies and I haven't met that many But there's always
- 25:08a few maybe who haven't really embraced performance marketing or digital marketing
- 25:13And so I have this deep kind of you know question for them, which is like why not?
- 25:17Why haven't you done so?
- 25:19And so my recommendation to them is like no matter what the reason is There is no
- 25:23valid reason today not to be in that space So you have to do that and then once you
- 25:28Get on the train, of course, then you're gonna get all of this data that you're
- 25:32talking about Uh, but if you don't like you rightly said it is a disadvantage
- 25:36you will be left behind and make no mistake Today's world it's not a choice.
- 25:42You have to embrace some element of performance marketing And because
- 25:46of that you then have to embrace data driven marketing analytics like
- 25:50mmm or mta You have to be able to measure it like Five years ago, I did
- 25:55a survey and we did a survey and we saw that something like 72 percent
- 25:58of marketing organizations or rather.
- 26:01A 70 percent of organizations that have a prominent marketing function
- 26:05are already adopting MMM and MTA.
- 26:06This is five years ago.
- 26:08Today, that number is going to be in the eighties, maybe in the nineties.
- 26:11So you don't want to be in that 10 percent that don't do that, right?
- 26:14Because that's just going to pull you back.
- 26:16You know, it's not regressive.
- 26:18Itir Eraslan: Yeah.
- 26:18I mean, like there's a lot of debate around performance marketing
- 26:21already, like the ad world and everything is just like, I mean, as
- 26:24a marketer, I, Sometimes question it.
- 26:27Although I understand that because it's measurable, you can see it and you
- 26:30can also segment it and those things.
- 26:33However, there's also like a big question mark around if online
- 26:37performance marketing is working for every brand at every stage.
- 26:41So that's why I think maybe, you know, when adopting AI, it It's can
- 26:45be, uh, good to see how much of the marketing business are we evaluating
- 26:51right now, depending on how much of advertising we are giving online.
- 26:54Uh, so if it's only like 20 percent of our budget is dedicated to performance,
- 26:58which is measurable, then the outcome that you expect from AI at this stage, uh, is.
- 27:05Representing the 20%, almost 20 percent of what you have been doing as a brand.
- 27:10So there will probably be something that won't be measurable, uh, in the very near
- 27:15future at least Uh for marketing and I think that will be like the debate that
- 27:20we have we we will keep on doing as a marketer Uh with the management teams.
- 27:25What did you do?
- 27:25And what's the ROI on what did you do?
- 27:28Uh, so there will be I think still a gray line, uh in the upcoming
- 27:31years That's my Personal prediction.
- 27:35Anindya Ghose: Yeah, I see where you're coming from.
- 27:36I mean, I guess what I'm saying is if a company or a marketer is willing, right,
- 27:42to adopt these toolkits and methodologies, the good news is they exist.
- 27:48They exist and they are progressively getting remarkably precise.
- 27:53Itir Eraslan: Yeah.
- 27:53Anindya Ghose: Now, the bad news is, as an outsider, I may not be able to convince
- 27:59the marketer that you should adopt, right?
- 28:02That becomes a very, sometimes a subjective decision, a cultural
- 28:06decision, you know, issue of inertia.
- 28:08So, um, in the book, uh, Ravi and I talk about the three
- 28:12I's, which are a big problem.
- 28:14Okay.
- 28:14So the first I is inertia.
- 28:16Okay.
- 28:17And the second I is like, like lack of imagination.
- 28:20The third one is like lack of innovation, but that comes afterward.
- 28:23Okay.
- 28:24But in our, uh, you know, in this conversation we're having now, I think
- 28:27this inertia, And potentially lack of imagination is very relevant to this
- 28:32context where I can bring the horse to the lake to drink the water, but
- 28:37if the horse doesn't want to drink it, You know, I can't solve the problem.
- 28:40So, and I definitely see that.
- 28:42I, I, I'm not saying that, you know, today there's going to be still enough
- 28:45people, uh, who often question, you know, the value of adopting these,
- 28:52uh, toolkits and methodologies.
- 28:54And
- 28:54Anindya Ghose: I think, you know, I always look at this as a conversation, right?
- 28:58My job is to kind of inform you and empower you.
- 29:02At the end of the day, it's the marketer is not willing to take that next step.
- 29:07And use the information and the empowerment then that's the most you
- 29:11can do like you can't really solve that problem So, uh, I guess the nutshell is
- 29:15the good news is we have the technologies and the models and the algorithm to
- 29:19solve the problem But the marketer has to first change their mindset.
- 29:24If at all, they're resisting, they're pushing their
- 29:25Itir Eraslan: mindset.
- 29:27Since we talked about performance marketing and how it's crucial
- 29:31for the brands to adapt soon because of all these aspects.
- 29:34And because of your, like you are the expert in this for so many years.
- 29:39In the era of AI, what traditional marketing principles do you believe still
- 29:44relevant for a marketing department?
- 29:46Anindya Ghose: Oh, sure.
- 29:47It's a really good question.
- 29:48Um, I would say I have maybe four or five I can think of.
- 29:51Uh, first of all, customer centric approach.
- 29:54I don't think that's changing, you know, understanding and addressing customer
- 29:58needs and pain points remains crucial.
- 30:00AI can enhance this by providing more deeper insights from your data analysis.
- 30:04But the core principle.
- 30:06Or prioritizing the customer's experience and satisfaction is timeless, right?
- 30:11Another one I can think of is brand trust and authenticity, okay?
- 30:14So building and maintaining trust through consistent authentic brand messaging is
- 30:18still important It's not going to go away In fact might become even more important
- 30:23because of the possibility of fake news that ai can also permeate, right?
- 30:28So some people will say well, uh, anindo doesn't ai personalize interactions,
- 30:33so it'll You know, solve this problem.
- 30:35Even if AI personalized interaction, I think the authenticity of the
- 30:38brand's voice is what resonates with consumers at a deep level.
- 30:42So I still think that's a fundamental principle that will not change.
- 30:45I think, you know, segmentation and targeting, right?
- 30:48Yes.
- 30:48AI can allow and enable more precise segmentation, but the, you know, the
- 30:53idea of segmenting the market into smaller units, telling messages, I
- 30:57think that's a cornerstone of marketing.
- 30:59That's not going to change.
- 31:01Um, and I think maybe lastly, I'll say relationship building, you know,
- 31:04I still believe that a lot of good work gets done through individual
- 31:09relationship between people, right?
- 31:11So I think, especially if you want to, like, nurture loyalty and retention, you
- 31:16know, uh, these toolkits are going to enable it, but at the end of the day,
- 31:20it's a human interaction, the one to one interaction, uh, that needs to be done.
- 31:26And AI can enhance that relationship building.
- 31:28But the underlying principle where you have to make the effort to engage with
- 31:32another human being that will not change
- 31:33Itir Eraslan: Uh, there was one thing that I want to ask one last thing from
- 31:38the book that I want to ask There is a word that I found which is quantified
- 31:43self I find it quite interesting because it's quantified and self.
- 31:48Can you explain that a bit?
- 31:49Anindya Ghose: Yeah, so this hits home because of my mountaineering hobby as
- 31:53well So okay quantified self refers to the practice of using the technology Like
- 31:58a wearable device, okay To collect data on various aspects of our everyday life
- 32:04like physical activity how many steps you walk, you know How many miles you
- 32:08run your sleeping patterns, you know, deep sleep rem sleep your nutrition and
- 32:12diet your biometrics because basically To be able to excel, uh, forget
- 32:17mountaineering just to be able to live a good fulfilling Uh life, you know, you
- 32:21need to be self aware of your individual, uh, physical parameters, right?
- 32:26And so what this wearable device, uh movement has done is it does help us
- 32:31quantify our wellness, you know, so nutrition sleep and exercise the three
- 32:36things now you can quantify it, right?
- 32:39Um, we have actually done a lot of research with you know, medical
- 32:42practitioners On in this space we have shown that roughly about a 15 month
- 32:47period after adopting a wearable device You The patient, for example, of chronic
- 32:52diseases, we see that, uh, patients with diabetes and blood sugar, they show a drop
- 32:59in their blood glucose and H1AC levels, uh, by following the three practices of,
- 33:05you know, uh, better nutrition, longer sleeping, and more exercise, okay?
- 33:10And we do these randomized experiments to monitor patients.
- 33:13Uh, behavior.
- 33:14So that's what the quantifier cell comes in.
- 33:17Um, and then, you know, in the book, we talk a lot about how AI
- 33:19is actually improved healthcare.
- 33:22So we talk about personalized medicine, you know, like, um, AI can analyze
- 33:26genetic information of patients and offer personalized treatment plans.
- 33:30We talked a lot about like predictive modeling of disease outbreaks.
- 33:34And the likelihood of a patient being readmitted to a hospital, uh, and one
- 33:39of my favorite example is actually from Hungary, uh, in, in Budapest.
- 33:42So, there's a, a series of breast cancer clinics in, in
- 33:45Budapest called Mama Clinica.
- 33:48Um, here, the doctors have been using AI to review and detect breast cancer
- 33:53diagnosis by the two radiologists.
- 33:56And, you know, they are human beings, but that's why they can make mistakes.
- 33:58They're experts.
- 33:59But like every expert, you can make a mistake.
- 34:02I don't think that sometimes the AI agrees with the radiologist,
- 34:04sometimes it disagrees.
- 34:06But, uh, in fact, there was a New York Times case on this as well,
- 34:09that in these marmosite clinics in Hungary, there have been 22 cases
- 34:14documented in the last three years.
- 34:17AI actually identified cancer that was missed by both the radiologists.
- 34:24Oh, okay.
- 34:24And these are
- 34:25Anindya Ghose: 20, yeah, 22 people now owe their lives so to ai, and there's
- 34:28another 40 people that is under review, so I don't have the latest results yet,
- 34:33but some significant fraction of that may also turn out to be a false negative.
- 34:37That means you're talking about 30 or 40, 50, 60 lives being saved.
- 34:42And that's just like in one clinic, in one city, in one country.
- 34:47Now extrapolate in multiple thousands of cities and, you know, 196 countries.
- 34:53Itir Eraslan: So I will ask a question about, um, about mobile economy, because
- 35:00your first book tab, uh, this, this is the last question about, uh, on this episode.
- 35:04And your first book tab was about mobile advertising and with now, with the rise
- 35:10of AI and all the other technologies, how do you see mobile and mobile
- 35:15advertising evolving in the next years?
- 35:18Anindya Ghose: Yeah.
- 35:19So I think it's again, a great question, which encompasses, you know, the previous
- 35:22question you had, which is are certain marketing principles still relevant
- 35:26and what will be different, right?
- 35:29So I think many of the principles we talked about will still be the same, but
- 35:33the something that will be different, especially in the mobile economy is
- 35:36again, content creation and curation, because now AI not only produces
- 35:41more interesting multimedia content, diverse content, but it does this at a.
- 35:46faster speed and at a bigger scale.
- 35:49I think with AI and mobile marketing, we can now map this
- 35:53customer journey a lot better.
- 35:55You know, remember I was saying that the traditional IDA path was linear
- 36:00awareness, interest, design, action.
- 36:02Yeah.
- 36:02Anindya Ghose: That's not how it exists anymore.
- 36:03And today in the mobile economy, you know, when more and more people are using
- 36:07mobile devices to finish their entire shopping journey, We have to do a much
- 36:12better job in mapping the changes in the previous journey to today's journey, which
- 36:17have become nonlinear and nonsequential.
- 36:19So I think, you know, mapping the customer journey, better content curation,
- 36:23speed of execution, and more scale.
- 36:26Uh, and lastly, I would say that I personally still believe
- 36:29that marketing is going to stay.
- 36:32a blend of creativeness and data driven analytics.
- 36:37I don't believe I will replace creativity and I don't think
- 36:40it should replace creativity.
- 36:41I think these two things are like two sides of the same coin.
- 36:45There should be a strong intertwined relationship.
- 36:48Uh, I think the job of the marketer is to think about the balance, right?
- 36:52What is the right balance between creativity and AI?
- 36:55So that data will actually inform creativity rather
- 36:59than stifling creativity.
- 37:02Itir Eraslan: Yeah.
- 37:02Yeah.
- 37:02On that side, when I think about the marketer's role and
- 37:05marketer jobs, uh, that's going to be available in the markets.
- 37:09Uh, you know, I see a lot of data engineers getting involved, uh, but The
- 37:14other part of it is that the current marketer or the upcoming marketer is
- 37:18getting, um, skilled, uh, in using and dealing with these technologies, not
- 37:24necessarily need to be being a data engineer on its own, uh, but knowing
- 37:29how to, uh, work with a data engineer or like a company with a data engineer.
- 37:35Because I mean, like I, as much as I read the upcoming jobs, for
- 37:39example, I was doing a search, uh, and it's all about data engineers,
- 37:43you know, cybersecurity and so on.
- 37:45And as a woman, for example, I'm just like, question.
- 37:47Okay.
- 37:48Uh, because I know that the data engineer roles are usually male driven.
- 37:52And I'm just like, how does that sit as a, you know, gender equity
- 37:55site, which is another episode, another conversation around that.
- 38:00Uh, but yeah, I feel that.
- 38:02A new marketer should be able to, they don't necessarily need to deal
- 38:06with data on their own, but they need to understand and capable of
- 38:09understanding how data works and how it can help them on marketing.
- 38:13Exactly.
- 38:15Anindya Ghose: Totally agree.
- 38:16Itir Eraslan: Two questions about lifestyle.
- 38:19Those are my closing questions.
- 38:21One of them is, uh, is there a book that you would recommend highly, uh,
- 38:25other than the book that's coming up, which I'm quite excited to read soon.
- 38:30Um, first of all first question is when is the book coming out?
- 38:33Anindya Ghose: Oh, so, uh, the book's coming out on october 1st
- 38:36It's available for pre order on amazon and some exciting news Uh, it
- 38:41actually hit three bestseller lists on amazon last week and the week before
- 38:45Itir Eraslan: Well, i've seen that i'm quite excited about that.
- 38:48Yeah.
- 38:48Anindya Ghose: Yeah
- 38:49Itir Eraslan: So first of october it's coming up and what's your favorite
- 38:53book in the meanwhile what we can read
- 38:56Anindya Ghose: yes, so This may be a surprise, but my favorite
- 38:59book has nothing to do with AI or technology or analytics or data.
- 39:04It's a book on mountaineering.
- 39:06Um, it's a book written by Ed Wister my favorite, uh, role models
- 39:11in mountaineering and the books called no shortcuts to the top.
- 39:16And the reason it's my favorite is, uh, it very much.
- 39:21Embodies the way I live my life or pursue my career, which is
- 39:25diligence, diligence, diligence.
- 39:27I have never believed that intelligence is as important as diligence.
- 39:32In other words, In my opinion, uh, whatever success metric, um, you know, one
- 39:37can evaluate anybody with including myself I would say 99 percent of my success
- 39:43has been diligence and non intelligence Meaning that you have to work hard.
- 39:48You have to burn the midnight oil.
- 39:50You have to be patient You have to study, you know, stay focused stay gritty, you
- 39:54know, uh, keep the mission in mind So the adversarious climbing style is very
- 39:59different from Someone like Steve House.
- 40:02They're both brilliant, like, you know, a world class mountaineer.
- 40:05Steve House does what we call fast ascents.
- 40:09Very light expedition backpack, 20 pounds, fast ascents.
- 40:13And at Vistars, does not believe in that.
- 40:15He does long, multi week ascents, heavy backpacks, heavy
- 40:20loads, slow, steady at a time.
- 40:23You know, nothing wrong with either one of them.
- 40:25It's a personal style.
- 40:27And the reason the book resonates with me is I believe that to be
- 40:30able to achieve whatever metric you have kept for yourself, you
- 40:33have to work hard, you have to stay patient, and you have to be diligent,
- 40:37tenacious, you have to persevere.
- 40:39So that's why the book resonates.
- 40:41Itir Eraslan: Perfect.
- 40:42I'll leave it in the links also, uh, as episode links.
- 40:45And what's your favorite place for coffee, uh, in New York City?
- 40:49Anindya Ghose: Oh, oh, that's a hard one.
- 40:50So on the weekends I go to Blue Bottle Coffee because I love their pour over.
- 40:56Itir Eraslan: Oh yeah.
- 40:57We can be good friends on that.
- 40:59Anindya Ghose: Okay.
- 40:59Okay.
- 41:00Great.
- 41:00Uh, yeah.
- 41:01There's Blue Bottle multiple places.
- 41:03You know, on a regular day, I think I'm always like stopped for
- 41:06time running from here to there.
- 41:07So I just grab maybe, you know, sometimes Starbucks, sometimes, well, there's
- 41:12a Felix Coffee in my neighborhood.
- 41:13I live in Met Town, so they're pretty nice.
- 41:16Uh, but sometimes it's Mato, sometimes it's Starbucks, so.
- 41:19But on the weekends, I make the 10 block walk to Blue Bottle Coffee and, uh, You
- 41:24know, it also helps me quantify myself.
- 41:26I put some steps,
- 41:27Itir Eraslan: which, which branch is that?
- 41:29Anindya Ghose: It says, uh, 19th and Park Avenue.
- 41:32There's a blue bottle coffee.
- 41:33Itir Eraslan: 19th.
- 41:34Okay.
- 41:35Okay.
- 41:35Um, my favorite is one of the ones in Tribeca and currently they have
- 41:39the Brundi, uh, single origin, which I, which I love a lot.
- 41:43Anindya Ghose: Yes.
- 41:44Both are awesome.
- 41:45Itir Eraslan: And I hope we can drink that together one day.
- 41:47Anindya Ghose: We should.
- 41:48We should.
- 41:48Definitely.
- 41:49Itir Eraslan: Thank you so much for joining me Anindo.
- 41:52Anindya Ghose: Thank you.
- 41:52Thanks for having
- 41:53me.