5. Conjoints and Coupons: The evolution of quantitative marketing

It's actually the next wave is actually going back to how you can make marketing decisions or targeting decisions with data. One of the very strong ones in it. And you will see this not just name be students but actually entrepreneurs. It's better when they say we are spending on marketing. What they actually mean is, we are spending on Advertising. This is called segmentation in marketing. It's also about cluster analysis in statistics and machine learning.

They are one the state. Hello, welcome to data, shatter the podcast on all things data. This podcast is a series of conversations with experts and Industry leaders in data. And each week. We aim to unpack a different compartment of the data suitcase. I am your host Catholic charity. That I'm a blogger newspaper. Columnist book, author, and a former data Etc. Consultant at currently head, analytics, and business intelligence for delivery. One of India's largest.

Just logistics companies. You can follow me on Twitter at Karthik s that is Kar thi KS and read my blog at. No Intruder.com. That is n 0e. N th you be a.com all opinions expressed in his podcast belong to me and my podcast guests. And it do not reflect the views of any organizations. We might be associated with nothing. Disgusting, this podcast should be taken as Financial or legal advice.

If you go to business school in India, you would end up with the impression that marketing is not a quantitative subject. Instead of weird process happens in the first deal. When you see a more mathematically inclined students gravitate towards finance and marketing gets labeled as for creatives donut. However, marketing has for a very long time, in an extremely quantitative discipline as the

amount of data explodes. It only seems to be getting more quantitative so that it's only right to get an assistant professor of marketing from I am angle to talk about marketing being a quantitative subject. My guest today is between us because G, he teaches courses on marketing management, and marketing research for MBA, students and doctoral students apart from this. He has a book title quantitative, marketing research.

Search available on edx and swipe his research interests include behavior and decision-making. Very models biases and digital marketing. When he investigates influence of Fraud and click. Delete. So you teach marketing at IIM Bangalore, you have done a mooc in, you teach a mooc and quantitative marketing. So, now based on 15 years back, back when I was a student at I am banged up.

There was a strong correlation between sort of people who are not so good at maths by MB standards and people who wanted to pursue marketing. It is almost like it is like if you don't do well at maths in in your first year in the NBA, then you are better off. Specializing in marketing and back then marketing is all about feeling and personality and things like that. But I understand that like there's a fairly long history of numbers and quantitative methods

in marketing. And since you also sort of work in that and teach a course on that, can you give us a sort of

a brief introduction on numbers in marketing and how it came about in for? Sure. Thanks, cryptic. So, let's start with what you said about feeling. And what is the other words? Personality, I do a lot of engineering elective in marketing. So yeah, so a lot of this actually, what when you talk about feelings and personality characteristics Etc, by the I just correct you. These are also these also have some quantitative basis. So let's start off with that. But I will go a little back.

Let's do a little bit of history as to how marketing is a field evolved. So it's all businesses are more or less derivatives of the so-called classical feelings. So most of marketing comes from econ with I think, you know, is a Quantrill. Most people will identify it with it as a cornfield psychology or consumer Behavior. Now, if your audience is primarily engineering, the probably don't know, but psychology is a intensively cornfield, and that's where the feelings and that's it.

So the problem of measurement is a quantitative problem. Sociology against astrology, has a lot of quantitative underpinnings, the entire ideas of network science. Etc. Lot of it comes from sociology. Yep. So the social networks that we see etcetera, which also has quantitative underpinnings and it's of course, statistics and operations research. This tends to be most of how marketing is the field evolved. So people from this field. So you may have heard of Philip.

Outlet is an economist, actually, another famous name Galilee. He comes from an operations research background. And even today, if you see marketing departments, at least academic marketing, departments hire psychologists, they hire sociologists. They hired a and of course, today, computer scientist, but historically, statisticians and operations research people. Okay, so you may want to ask where numbers? Come in actually numbers. Come in almost everywhere.

It is a small aspect of marketing called qualitative inference, but it's not just about your intuition. It is a rigorous field in itself. It's something that I do not specialize in so I will not go into. So let's go into the quantitative. Underpin. At the hut. A very reductionist approach to marketing is what that you need to make something and sell it for money. Yeah, that's the most reductionist idea of marketing. Of course, that is not entirely

true, there's more to marketing. You can be a non-profit. You may want to have more nuanced idea of marketing is that you have something that you want to exchange with somebody else for something else. Yeah, so you may want to develop an impression of yourself and then go certain actions Etc. That would also come under the Ambit of marketing. Let's go to the most reduction is but the one that most of us can understand that you have to iron Sell you to buy and sell

right now. So, here, I'll use some technical terms. So the idea is that you have a product and you want somebody to buy it. Yeah. Well, you know, that not everybody is going to buy what you sell. This is simply because different people have different needs. Yep. All of these products that are selling our services. They address some certain need, and why is the other person going to pay for it? Because the amount of money here she is going to part with Is.

Worth less than the benefits that he or she receives from your product product. Yeah. So add the heart of it is an exchange is a profit loss calculation. That itself is quantitative. Yes, right. So initially, let us say the 1950 which are considered some of the earlier days post wash. There's a boom that but you don't have enough data to know who wants one because it's you don't have databases difficult to track a person. What do you do with at the store level? You know, that today?

I sold certain things at certain prices and I observed profit plus Etc. Now that storekeeper may increase prices decrease prices Etc. And that is the rough proxy fuzzy price elasticity. Now that is not completely accurate because he does know who. But what me as a person is very different from you as a person or a third party the person

Later what Market has devised is something called the coupon direct mail. This at least when I was growing up. This was not very popular in India, but apparently in the western economies u.s. Especially the coupon was a very, you are. Do you say way of tracking, who was buying, what you find coupons in your mail? And you can go get discounts or whatever else. But based on that, what ever you have redeemed? They're actually keeping a

database. And then the next coupon you get is going to be very different from the next group on somebody else gets. So essentially they are using some sort of statistical modeling. Track you in a way and go deeper and deeper into understanding. What your preferences are. Now, of course, coupons are expensive. It you have to pay postage having a thinking that there be in India have sort of Leap Frog

coupons. We leap from Cooper really did get coupons, but I do not remember getting that they take referrals started Cooper. The first time I saw targeted coupons was actually in France in a supermarket. So I have a loyalty card. This is this is last decade the right thing, but I'm afraid. It still goes ahead. I have a loyalty card. After I make my purchase, I get a coupon which is very different from the coupon that the next

person. In my, in the line, gets my coupons are somehow eliminated based on my purchase Behavior. So, coupons exists to this day. Of course, there is a for the inertia or language or you call it. The original idea of coupons was that you sent posted? Yes. This is still inefficient expensive and all of that. So the what happened? The next wave? And this game in the 1980s. And again, we have kind of leaps of this. We are in this scenario to some extent, the idea of Supermarket

scanner panel data. Okay? Yeah. Now, scanners panel data will tell you who bought what they can track you based on your library card. They can track you based on your credit card Etc. You are not going to get everybody because I can still walk into a store. I can be a walk-in. Yep, take cash and get out. So they don't know where. But a large number of people in the western economies were using credit cards, of course, right? So they can track you with the credit card.

They can track you. If you have a loyalty card with, where they give you a discount, Etc, today. However is very interesting. Amazon doesn't need to give you a ticket because they have all your details in. Yeah, you have your login ID, so it's slick. Yeah, so literally, you can track every customer and their purchase history. Yep, essentially. Of inference is how you can use data of the past. To make decisions. And when I say decisions, I

mean, can I tell a product? So you can you tell her offers for you? Can I tailor ads with yeah, right. The same thing. If you take a look at advertising, you have TV ads and you have been assets, you have newspaper ads which are not really track you. Of course you have cookie ads online ads, or you have mobile ads which can track you. Yes, so individual. Double date actually increases and this is why actually Google became the multibillion-dollar company, it is right? And Facebook.

Yep, because if you remember pre-google, there was no way to find out who you are. What your preferences are and Target you. Yeah. But to some extent you can so I can say that. Yes, if you are in a posh area, Supermarket. Yeah, then maybe your preferences are going to be a little different from if you're not in a posh area, tailbone down to your foot down, Etc. But it still approximate on Google. However, you are searching for

exactly. Something. Yep. They have a much better idea of what your preferences are and they can use that information. So now that has been the evolution of how quantitative data has been used. So previously people, you should do a lot of surveys, Etc. They still do today. However, you can actually find out what people want. So, surveys are essentially what I tell you what. I what I tell you may or may not be true all the time. You asked me? Do you like this?

Yeah. I miss, you know, let's say it was exactly that Google knows XYZ about me. But doesn't like resting this evil is Yes, are accusing another classic coming in the literature. They call it as like stated and revealed preference. Right? Like we used to trade another based on stated preferences until now, but now we have enough data to kind of like, we don't need to know your status Bristol. Yes, for a lot of things, you know, you're you know, revealed preference is not for

everything. Yeah, but for a lot of things now, however interestingly this is GDP, our rule that's come up in Europe where you are not allowed to store certain kinds of data. Yep. So, if I were to take the times of India, homepage in India, versus the same time, the finding home page in Europe. Let's say in Paris. I would have to authorize a lot more gas. I have to obtain and most people don't while here automatically a lot of stuff is being recorded.

Yeah, it is true for any online player. According to European Lungi repair less this happens. So actually the next wave is actually going back to how you can make marketing decisions or targeting decisions with data. Yeah, we were going up. And again, there is a wave of at least in certain economies where limited data is Rule. Yeah. Okay. Thanks for the demolition. I mean, like this is used.

So what I understand from you is that like marketing is always been a highly quantitative subject and like you have people from like sociology economics, psychology statistics and so on which are all highly quantitative discipline.

So why is it that we in India or at least in is we think? Think that we think that marketing is a soft subject without much Quant. Why is it that we have this weenie? Okay, that's an interesting question. There are parts of it, which I will not answer why you in irons. Think something is something. I'm not qualified to answer. Okay, that's Michael. And I have never been an iron

student. I've only seen it from one side by exposure, to marketing has been from a European schoolnet was pretty cool and heavy so I actually did not have the same Notions that you did. That said there are certain marketing role. This. That add the face of it. Do not seem very pointed and this is absolutely true if you're working for Unilever and they send you, this is rap to rural beer. Yeah. That, that will be an initial

training placed right now. Of course, I think everybody has to do it, especially in the top of Energy company. What if you're doing B2B sales, where you have to negotiated deal with a buyer Prime, see a lot of these marketing change. Now, the pills does not see until you can only guess that this is probably one. Is driving the perception. However, at the back end you have to maintain says, Diaries salesforce.com is a huge business.

So is as it is associate, switch does things like calibrating Salesforce incentives, and training Salesforce people Etc. And the back end. There is the lot of machine learning. There's a lot of statistical analysis, economic analysis, even psychologically. So at the end of the day negotiation is this type is a psychological process. At the back end, I did big companies at the back. End will be keeping tabs of a lot of data. Yeah, in B2B settings. They do. Keep that record, sales persons

died. Is it is to be die, least lead generation in cetera. Nowadays. There are software systems that track at what level the customer is. If you are talking to the customer, what exactly happened is the customer a On to the next level. Then they try to do some analysis, regression analysis or something higher trying to either segment customers into different groups or trying to understand why, or why you did not get a deal, etc. So a primer facie right in the front end.

It may not seem to be very quantitative. But at the back end, there is a lot of quantitative analysis going on. Similarly, with banks at the front end. You have some guy calling you up and pitching an investment product or whatever or credit card or whatever. But at the backing they're tracking your transactions. They're tracking how much money you have balance, Etc, and trying to make some sort of influences in particular. So, okay.

So you're saying that. Like, while we might think that it's Bit of a sort of non-quantitative sort of discipline. There is a lot of quantitative stuff. Just that people are not exposed to it in the earlier, not necessarily exposed to it in the earlier years of the jobs include cigarette. Earlier years of the jobs in, the second Factor will be probably the core course in marketing that started in. I am at least I am Bangalore, is not very quantity.

Me. Now. Remember you are bringing in people actually from an engineering background. Very few people have actually exposure to marketing. Yes, and you have to teach them a lot of stuff to. One of the things that we find is that there's a preconceived notion and my marketing is equal to advertising. And a lot of people have this notion even experienced. People have this notion that marketing means, you're going to be the next person.

Joshi ZJ you spawned a huge point is no, the old gray. It is. Alec presidency. Yeah, so not a lot, but quite a few people tend to have certain preconceived notions, right? One of the very strong ones. In the end. You will see this. Not just name be students, but actually entrepreneurs. It's better when they say we are spending on marketing. What they actually mean is, we are spending on Advertising. Yep. Yeah.

Product designed. So you need to convince somebody that this product design, which was also part of marketing that is Retail format retail store designed choosing the retail format and things about the place. The 40's essentially, which is also marketing pricing is also marketing, Etc. So I think there's a lot to unpack in the first marketing course that people take the S marketing quiz that people take. If anybody is interested is

usually during the search. We This, you do end up doing a lot of construction. Like this is, you probably did not do that in your vad problem. It's an elective at. I am back and I did it and ended up with a backdrop. Let's not take names here. But like I did and they ended up with a profit made me lose all of my critics. Oh, yeah. Well, okay. That's that's beyond the scope of this discussion. Yeah, that's beyond this little marketing. Yeah, so marketing these hatches.

You start doing very shallow dive into quantitative methods, but there's a lot more actually than what is typically taught in a course. Okay, and it took us to sort of before we sort of dive into the actual contents. Only one other question. Is that like so everything.

I mean you are an academic. So everything that you have told me probably comes from an academic perspective in the in terms of like how Quant is being used in marketing, how it's how it can be used how it has been used over the years but What's

the typical translation like of what happens in Academia to Industry? Like, sort of like, is the from the industry standpoint also apart from your digital marketing and stuff, which is still very quantitative, or is it sort of like, very different from what it is, in Academia? Okay, so whatever I told you it. So I have a purpose, my experiences are colored by the By the academic good point.

But whatever. I told you that actually use a lot in Industry. So for example, when you want to design a product, let's say you want to design a hotel room. Okay? Now ideally you would like the best most luxurious hotel room at the lowest price. Giving free Wi-Fi, giving like screen TV, whatever. It's I access whatever. That's not feasible. It. What you want to do is kind of try to measure for example. For a given price how prices it is? A customer is of a singer.

But along with that. Where do you get your most bang for buck? Okay. Should I give free spy Axis, or is it better to be free Wi-Fi? Which one do I think will lead to more customer satisfaction or bring in more customers? Right? Ideally, I would like to give birth but I don't want to because the cost money. So there is a method called stick conjoint analysis. For example, where They say that this hypothetical. Sorry, let's dive into conjoint analysis. Now that you have have let you

do it alone. Yeah, okay. So this is just one example of where industry actually uses quantitative methods. So what? So what happens then is that? I will give a survey. To a customer or potential customer. That gives him a set of him or her a set of hypothetical hotel rooms and I'm purposely using hotel rooms here because the Hilton group has very same. Mostly used conjoint analysis. I think 30, 40 years ago to design its various offerings, 30 40 years ago, ha ha It's an old method.

Okay, well worth it. So I give you. So, for example, I say 10,000 rupees the night, free spy axis. No internet, okay, option one, another hypotheticals. And now you can see the more so-called attributes. I give I can actually have hundreds and thousands and even millions of potential combinations of permutation combination problem, right? So I can change the price for anything. I want Wi-Fi. I can say free Wi-Fi or paid Wi-Fi, but I can also increase

that to 1GB a day. 25gb a death certificate. Yes, all of them cost money Etc. I can have spy axis. I can have free breakfast on. Not wreckage. Very the price of breakfast, etc. Etc. So I can give you, hypothetical scenarios like this. This room has a king-size bed free Wi-Fi. No Spa access option. One option, two. It is 8000 rupees. 35 axis, free Wi-Fi, Etc. In the process. What I am trying to do is I am trying to find out the consumer

or the customers value system. What you value this, what if it doesn't. So some then maybe a group of customers that just wants a sea view, you know, time in Bombay. While another guy doesn't care about this review, but once free Wi-Fi, yep, maybe maybe. He's on a business trip. You want sleeve is like the third guy. Maybe he doesn't like either of these two, but would prefer free breakfast. Yep.

Stuff like this. So what I will try to do is figure out what customers want or what groups of customers are yet accordingly. I will Design my rooms. Yeah. Now for the same price, I might give customer x a City View. Rules. Because he doesn't care about the Sea View, but free breakfast yet because we might prefers. The breakfast were at the same price. Well, another, I will say, I don't care about free breakfast. Happy to pay for basic first, but at the same price, I want a TV room.

So this is what conjoint analysis does. Conjoint actually means considered jointly multiple attributes considered jointly. Take a look at cell phone design. Yeah. It's a classic. Listen, conjoint. Analysis comes in. Do you want more battery life, or do you want more RAM in your phone? They are at the same price and you are price sensitive or not yet. But that price goes with Allure. Am sensitive or not. Are you screen size sensitive or

not? Etc. So I will give you hypothetical Community. I cannot be all the combinations. There are statistical techniques. That let me reduce the number of hypothetical scenarios. I give you and in the process, I try to figure out what you want the most. Megan's take my customers, bitch. Yeah, okay, that I was coming to

that. Like if you can also cluster and sort of segments to people because like they will do a bunch of people who may be willing willing to pay 2,000 Rupees per night, except for the Sea View, while another Bunch who might there, not value it at all, in things like that. So because exactly, Exactly. So, ideally what I would like is I would like to give you a room that exactly fits your specification and maximizes my product. That's not going to be possible. Yes.

I have a finite set of rooms. Of course. I have a finite set of rooms. So what I want to see is, can I find clusters of people? Yeah, who value C View and breakfast and this. Similarly, so can I have a, this is called segmentation in marketing. It's also about cluster analysis in Statistics and machine learning. They are one in the same. Actually. Can I find a group of people who have similar needs or preferences? And then I will give them something where they are.

Where they're more likely to like my product. Yeah. And healthy make more money at the end of the day. It's today made for money. Yeah, so I guess identity for phones. It can be for false. It can be for hair saloons.

It can be for IT consulting. Contracts. It can be for hospitals, can be any of this. But I'm sober. So tell me anything like that. So I think conjoint analysis since its 30. 40 years old. I'm assume assuming it was initially designed for the sampling world where you do a survey and then do this analysis. Now, I guess, like, if you are, if you are, let's say I make my trip you.

Possibly have the data that taking your hotel room example, you probably have the data to actually do this with what I would call as population data, rather than sample data as that actually happened as in like as and are has marketing evolved to sort of, take into account, all this the amount of data that we sort of collect and use nowadays. That's an interesting question. The answer can best be given by

companies who have this data. It's very unlikely that they're going to tell you what exactly they're doing with it. Of course, on my guess is yes. My guess is that there must be some sort of background analysis. Done travel booking websites, are a classic case of where you can take. So you cannot exactly. Do conjoint analysis, which travel booking data, by the way, organically because I have given you only one choice.

You have given me certain options which actually exists and I've given you one choice and your clothes. You don't know my relative references for XYZ, over a large number of people, you can do some sort of statistical inference, but at the individual level, I have given you very few data points, so But yes, what you can do is you can a find out across a large area, almost population level.

What people's preferences are, you can also track me and my past purchases assuming that I've been loyal to make my trip. Now, that if you were taking the example of make my trip, now, that is not necessarily true, of course, along with make my trip. We'll be able to clear trip. I may be going to booking.com. Yeah, and a lot of other websites that actually takes me

to another very interesting problem in marketing. Okay, which is, which is modeling vertical, steamer attrition and Customer Loyalty. Okay. It's a very, very big problem in bucket. Okay. Let me give an example. Let's give you the example of make my trip itself. Yeah, so, okay. We'll go back to for Simplicity sake. Let's go back up. Recovery data, okay. Okay. Yep. It complicates a lot of things. Yeah, we don't yet know how the models are to think, that's it.

Let's go back to the precordillera. So, make my trip has my phone number. Yeah, so they know. It's me when I'm booking also. Actually, when I booked a flight ticket, they have all my details, so, they know exactly. That it's me terribly. Yes. Let's say I travel on 1st January and make my first ever transaction one for Jen. 2010. Okay, one month later. I make one more trouble for months later. They make my next travel. One month later. I made it my next travel, Etc.

So they're seeing a pattern, right? Now after that, sadly they say, see a break or six months. Yes, okay, and they don't know the future at that point. It's been six months since I last traveled. So the question I have not signed any contract with me by trip. It's not like my mobile. It's not like a postpaid mobile phone, operator. Yes, I signed a contract and they know when I have pulled out the question is, am I still active on make my trip or not?

Yeah. Now you can think of it's a make my trip against Amazon, then snip it buzzer. Yes, is it this is called, I am a non-contractual customer. Has been frequently the store and at each time, they know how much I have but how much is spent it? So, now the question is, how do they know when I will be back next? If at all, I will be back. Yep, very tough problem to say. Yep, a first-rate here.

You can do things. Like, you can assume that if I'm in active for six months M1. Yep, but is it really true? Is there a better method? One of the things, one of the, this kind of modeling was started in the 1980s early 80s, okay. Buy some days called David Schmidt line. MIT guys, anything they wish with line Colombo is the third day Morris and Donald Morrison. How can I forget Donald Morrison? So what then suggested is that these patterns that you see are

actually a person process. So I am like a, yeah, I enter purchase times. Our poisson process are a person. I'm assuming you're actually in the processing will be a connection. Interposition stammer, his exponential, took its time would be exponential. Okay, exactly. Yeah, except at some point. There isn't stopped. Maybe I got the stuff. Yes. Be a located out yet. Maybe I just don't want to travel anymore. Or maybe I actually died. Yes, follow the first possible. Yes.

And in this literature is the estimate. This is actually called Death. It's quarter after my death. So Ellie What they came up with something called which is one of the bike till you die murder. Okay. I am a customer with a certain exponential distribution. That models my enterprise system. Yes, but there's also an exponential distribution and it's like a time to failure in a way, which bottles my lifetime. Exactly. Yes. There are two things that are simultaneously in done.

So based on their past purchases, you will actually have a likelihood function. Oh, there is a second step to this. Now. There are thousands of other customers, right? Yes. Who may not have the same. Exponential distribution parameters that I do. Yes, on top of this you need to have a distribution that models the heterogeneity of this, right? If you have to have fun is my Lambda 2 Lambda mu 1 is from a lifetime and one is from in to possess.

Yeah. But you and me and thousand other people, our same parameters, then come from. Let's say beta distribution. All are my distribution, depending on Etc, depending on certain assumptions. You can actually change that. So there's a lot of research going on here to use a distribution instead of an exponential distribution Etc, to model in to purchase type. The simplest is it approaches the times you see, exponential

distributions? And so they came up with a very important result which says that I don't need to store every single But just timing know it. Remember databases are expensive in the 80s in the 80s. It was exist. I thought I did if I use even now even now by the way, are even now even if you were to store everything and not use an exponential distribution and try to calculate a likelihood function, your computer is where even the very most powerful computers. They run for weeks possibly have

to what their true. So again, if you are you will probably know you don't know if your leadership sorry. Isness. Your listeners, the exponential distribution has a memory less property. Of course. Yes, which means there are three parameters that you need to stir. Yes, the timing of the last purchase, which one is the recency. Yeah, how many times you have purchased in the past, which is known as frequency and how much

money you have spent? Yes. It is. The last key, I think in the last or maybe the monetary transaction, you may need. The money part, I remember what the recency and frequency. Yet are sufficient to start modeling your future medium. Yeah. I got an aggregate level. I can totally believe that at at an aggregate limit and So further from people, simplified this Etc, these at an aggregate level are excellent in predicting future volumes of purchases, from a knot at an individual level.

Yes, but If you have a cohort of customers, it is very good at predicting how much they're going to buy in the future. Yep. Now, the other advantage of this person processes You Can Count. Yes, there is actually a method. There is actually a closed form solution where you can count the number of transactions in any given time period, what they show is that out-of-sample predictions are amazingly good for this. What is not good? However, is how do you tracks an individual?

So the next level is can you make individual estimates of Lambda, get a little distribution parameters, etcetera at the individual level that there has been mixed success, but people do that a lot using using hierarchical Bayes kind of analysis. Very computationally expensive though. If you try to go at the individual, Level things get a lot more complicated. And of course the next step is then at what point do I send you a coupon? What kind of poop and you send

you? How do you react to my marketing messages at the individual level? So this is a lot of where it's called customer lifetime value modeling sometimes or customer churn modeling. Yep, another very important. I would say quantitative area of marketing analytics. Yeah. Okay. I think I think by this point, Point. I think if anybody has risen done till now, any doubts, they have about marketing being a quantitative description would have been dispelled.

I mean, just this one thing of having to talk about customer lifetime value, which seems like a sort of, I mean, sort of a global concept. We've kind of broken with down to a exponential distribution and parameters such as like

frequency and decency and so on. And so it's very clear about how how quantitative marketing is. I don't think anybody will have a doubt about it. So why do I say so slightly switching tracks? So in the last 10 years, I think one new career that has come up, which wasn't there before? Was this whole thing is a thing called digital marketing. And what do you find? Especially when I go around on LinkedIn is that like sort of the people who call the age of

digital marketers? Have they don't seem like. So they sort of seem disjoint from traditional Market is, why is it that digital marketing? In your opinion has sort of grown in a completely different way and like, from from the traditional Market, Because the way I see it, it's all marketing. It's just a different medium. Okay. It depends on how you define digital marketing. I would say the pioneer of digital marketing comes from mobile essentially, because they

pioneered. And these are not mbas or even business activities in any sense. These are computer scientists who, who realize that I can find data about you. I can get data about you and I will then pass on that data. Our because to advertisers, you can Target you with with something very specific.

So, for example, if I look for, let's say, but our shoes I want to put on Google. Then Google will pass on the message to not just butter, but pursue retailer, or maybe Nike or maybe Puma. Yeah, when we're competitors to butter. And so It is better than sending me a leaflet on in my mailbox, where they don't know if I'm actually looking for shoes or not. You have the fit between the person and the advertisers is established. The next level is probably at

this point. The typical digital marketer that you may be talking about is the so-called social media marketer, you know, yeah, which is doing social media campaigns. And the so-called influences, where it is, somebody with a large social media following. Who is then missing whose tent going and telling the brand. Then? Look, I have so many followers. So, I am a very popular food blogger. Yeah, if you are selling The nonstick cookware. What I, I will.

Endorse it, and there's a benefit to you. So you pay me for it. Yep. Social media marketing is an extension of what Google Pioneer. Google tried a lot to capture social media, but for some reason they would not all of these are now. Yep. What bviously future? Nobody can say right now. Social media is a strange Beast because Social media lets, you know, not just what your preferences are, what you share. Our reference is economics, if I move on to the XYZ restaurant

and I put a picture then. They have some idea. What kind of Cuisine zelig etcetera. The fun part about social media is that it can map your social network for certain degree? Yes. It knows who your friends are. We'll see your posts. Yes, what there. And this is where actually it. So I told you about sociology yet. This is where social logical Concepts become, very important Network science, of course not. So, now Facebook and do

something that Google cannot. It knows not just about you but a body of Rex. Yes. Yes, and essentially, that is the reason why if Facebook is so valuable, this guide, the fact that you're not searching for information on Facebook also. Also, of course, Now, you have killing a lot of time on Facebook, Twitter, Instagram, Etc, which means that you have in this sticky on this platform, which means by definition, there are 24 hours a day. So your DV time has now become okay?

Yes. Okay, for example, a there are two aspects to it one is they have granular data. The second is it is in their interest to people who come to the platform because Only if a brand gets eyeballs, will advertisement this network, right? So now, Digital campaigns. In that sense, a little different from a flight Camp. It's little mention Machinery. Any to get your view on this. How has AI? Ml Big Data Cloud. How has that changed marketing?

Well, not right? Okay. Look at how Google is in marketing. That's your answer a lot. So if you use your data in a rigorous fashion economically residence, okay, let's be clear is it is scientifically rigorous. Let's say fashion. There's a lot that can be done with big data. So this for example, I started off by telling you that store inventory right in the face, too. Is Amazon knows what you clicked? When you abandoned your cut

shopping cart? Yes, they can do pricing experiments into price discrimination experiments where they keep one person. One tries on the person on the pricing, Etc. Yep. You can do large A/B tests. They can do a meters from their format Etc. And by the way, not just online even offline, big store can do it. And record the results are in a database. So it has revolutionized marketing, a lot, the more data you have. Oh, the more relevant data that

you have to keep the board. Informed decisions. You can make the trick is in how you use that data. So, I'll give you a word of caution. Same way that it has changed Marketing in a lot of ways. There is a lot of improper usage as well. Okay, so a classic example is, let's say you are modeling my behavior. And you have a 1 million meter model, OKAY? In your opinion. Is that a good model that a bad Body by Body? Is that a good point? A little bit further Back Body.

Why to my parameters, too many parameters. I mean like, it's just lucky to have a lot of articles in the Press actually said that all. Look there. millions of data points and they're using millions of parameters to A Target, you to predict your behaviors and you'll find a lot of people actually say that. Look, I can predict your behavior etcetera with this. The problem is a serious. Statistician will tell you, then. This is a very permanent.

So at the heart of the matter, it's still the same. The principles of Statistics are still the same way, you need that. You need pasta money and you need some sort of. Critical justification for the model pure data. If you were just to fix ourselves into a data, it's very unlikely to have out-of-sample predictive validity. The principles are the same, the number of data points increase, do you? So the statistical power of what you do?

Goes up a great deal as you mean that you're using the data appropriately. So, yes, in short the data, the greater availability of data. If used by the right person, in the right way, can help a lot. It can also be misused. And mislead you into thinking that you are. You have very sophisticated analytics which may not always be the case. Now, by the way, I don't normally plugs and be but I am be actually has started a new MBA called Indian analytics. It's it is an analytics focused

MBA. Where's a lot of your coursework is different from the standard. India does, if you can intake of only 40 people in the first batch, they will be graduating next year with the intention. Of course, when I say MBA in analytics, it doesn't mean MBA in marketing analytics. Yeah, let's go operations analytics HR analytics and all of that Financial analytics and marketing analytics actually placed a large. Parole today in the so-called

analytics. Boom. That fits in yep, whether it's music miles of the world or earlier, the Googles and absence Facebook's Etc. Other special genetics company, Sigma a lot of others. A lot of them actually look marketing analytics. So MBA is across the world are kind of you to live in this way.

I am be, has an analytics and behold a few American and European. Schools have specialized Masters or MBA programs that tend to focus more on analytics than a traditional lb Etc. Thank you for listening to data shatter. If you like this show, please leave a comment, share and subscribe to the podcast. You can find this podcast on Apple podcasts Spotify. By or wherever else, you go to get your podcasts. Once again, this is Karthik signing off. Thank you.