 Good morning everyone, this is the first time I'm speaking at a conference in Bangalore and glad to be here but before I get started just a quick check on what kind of backgrounds do all of you come from how many of you are working in some form of technology, would you just raise your hands how many of you are playing a pure business role, no technology out there and how many of you are working in analytics, so technology in analytics for you is the same or similar and I see a lot of hands coming up there, great, so that's a nice point to be so you know I have 40 odd minutes to speak today and what I really want to spend some time talking about is you know analytics often is a lot like the old story about the blind man and the elephant right the many blind men and the elephant and everybody seems to have a different take on what constitutes analytics there could be someone who does a lot of good business intelligence and that's analytics someone who does a deep algorithm and that's analytics and someone from the business perspective also sees it as analytics so you know often one of the things that happens when you have an unclear definition is you do get into paralysis analysis is paralysis is an old statement that all of you would have heard and from a business point of view if you're you know when you're working in business any form of paralysis is actually anathema you just don't want it and therefore that starts to become a problem with the function itself that are you over analyzing, are you actually getting into a paralysis point so one of the things that I've always believed is analytics is actually occupying an intersection set of skills so you need people who have technology backgrounds people who have deep business understanding and people who understand stats all coming together not necessarily that they have the same percentage of all these three together but the team which gets constituted needs to have all of this and when it's a question of bringing intersection skills together the challenge always is to understand each other's language so I have a geek in my team who is not understood by the guy who is a campaign manager who needs to take action on the insight who is not understood by the business guy who doesn't understand the tool so how do you start improving communication in the case of a business or a function where without that communication the intersection does not happen is absolutely key and which is really what brings me to the point that if you do not want paralysis to happen you need as functional specialist to start telling stories about the data that you see which capture people's attention and why that becomes particularly important is that one is the communication within your own themes but more importantly anything that you do in analytics has to go out there and make a difference to the business and if it has to make a difference to the business you have to hold their attention and that is where the concept of saying that how can we as the community of people who work with data with analytics whatever you might want to call it how can we become better storytellers and actually enable the usage of analytics that much more so what I'm going to do is I'm going to spend some time talking about a little bit of context and that context is really more to focus in on one aspect which is more marketing my own background is that I come from marketing for the last 20 years I was the CMO for SGFC bank till about 5 years ago and before that I was the CMO for a large retailer called Shopper Stop and so therefore my focus has largely been marketing and over the last 5 years I along with my partner Swamy we've set up a venture called security in partnership with the company called Hansa so the first part of my presentation I will take about 15 minutes on that is really more setting the context about the marketing aspect and how analytics plays a role and really what aspects one needs to cover there and then I'm going to take a small example of how data can be used interestingly to tell a story and then through that converge to say what kind of learnings in analytics should one carry forward into business that's really what I want to cover so just starting out I think we heard in the earlier presentation about the business whole tsunami of data lots and lots of data out there and in my view as we go along and as business grows I think for years and years together businesses have seen brands as being very critical and brand equity for most businesses has been very very important I think in the days to come data equity will occupy even a larger space in the minds of people within business and consumers as well and data equity if used effectively can translate to very powerful brand equity as well and I'll talk a little about that but how companies start to store use transform data can actually be a huge differentiator we hear about this in a lot of case studies how many of you have heard of a bank called Capital One right how many of you have heard of a casino company called Haras Casino right so at the same time you know that there are only these four or five companies which have created compelling differentiated advantage by using data effectively right so why is that if data is going to become so important and there is data equity which is so critical why is it that for that to happen you need to do more than just do the technical stuff on the data and I'll touch upon that as I go along but the point really is that if that data is not powerfully used to differentiate for you as a consumer your experience when you walk into a store when you walk into a branch when you go to a travel company if that difference is not felt by you as a consumer that data is not worth its while and over a period of time that data will get eliminated by businesses if they aren't able to find ways to utilize that so the final point that I'm making on this chart is to say that data equity while it is going to get very very critical not many players not every player not every company will find ways to transform the data and take it to the point of experience the point at which you as consumers you and me as consumers experience that company and that is absolutely critical so if I look at India for example you have more than 6,400 GDP now coming from services businesses so banks, retailers and so on and what that is doing is obviously creating far larger interactions for each of us with the retailers with the banks and that's producing data so all of us are actually leaving footprints behind of data with each of these companies which is up to the company to analyze and get insight from but that footprint is there the other piece which is happening in India again is that thanks to the growth of telecom and all of these businesses a lot of us in fact more than 400 million people in this country are today addressable what I mean by addressable is you are on a database you sit either in a bank's database or in a retail company's database or in a travel company's database or a telecom player's database and when you become addressable you can start to interact with you and you aren't just an amorphous mass of consumers so that's the other thing and obviously as this data starts to embed itself within companies there's the opportunity for companies to start having conversations with consumers and one-to-one conversations but at scale and a lot of companies like Capital One, Haraj have actually done that extremely well and that's going to be really the future where companies start building those conversations at the same time and I touched upon this earlier when I said that there are not that many companies which are building that compelling advantage so while the data is exploding the execution gap of companies being able to utilize the data is creating a very large void and then that gap really is largely about how do you create the downstream pieces of which analytics as a function is one part to enable people to take action on that data so when I am a credit card holder and I get a call from the bank which actually says that have you just purchased something which is high value if I walk into a jeweler's shop and I buy something of high value how many of you have experienced this from a bank? just a few, right? so that's the example despite so many banks and almost 25 million credit cards in the country just a few of you have experienced this which could either mean that not a lot of you are using the cards for high value but the other point really would be that not that many banks are connecting with you at the point when a high value transaction happens but when a high value transaction happens and I am using my card and the bank calls me saying that sir have you used your card for this jeweler's shop what really is happening is that data equity through the transaction processing system is coming to me and I am feeling a sense of comfort that this bank cares about the security of my card so if you remember earlier I spoke about how data equity can powerfully transform the experience and actually get brand equity at a notch above that's an example of how that can happen and then when you take this across to multiple businesses multiple industries you have the potential to really really impact business and then how do you get companies to do this because in many businesses this kind of stuff has happened because of the nature of the business the banking and financial services business for years has done stuff like this but how do you do that in consumer goods if I go to a consumer, if I go to a grocery shop and I am buying I don't get a set of relevant offers for me so that is the challenge which I think all of us as professionals have to work towards so just changing context a little bit you know the question really is how do we as consumers take decisions more often than not, we use our instinct yes we might use data, our day jobs might be saying that data analytics but in our personal life we use a lot of intuition and the other side as our intuition sometimes misfires we start to calculate, analyze and do more of that and that's how we start moving in that continuum towards being a little more analysis based sorry I am having a problem with this so the issue really is as one starts to look at this kind of analysis the issue to my mind is that is analytics really about the sexy insight that you can draw out from the data that fantastic diapers and beer example which one hears about so much have all of you heard of this diapers and beer example how many of you have heard, yeah okay so is it really about those sexy insights or is it about an effective decision is it about the fact that when I am at the store I get something very relevant for me I think banking and financial services have actually mastered that aspect if you go and apply for a loan in a bank the bank uses tremendous amounts of data and actually are able to do a go no go decision whether to give you a loan or not give you a loan that's an example or if you take a credit card and depending on how you repay and how much you spend your credit card limits would get enhanced so banks have done that very well but how often do you see this happening with other businesses not as much and as competitive intensity starts to increase as more and more companies need to start making very relevant propositions to consumers what consumers are saying is that if you don't make a relevant proposition for me I'm gone, I'm out of here so you need to become relevant so at one side this is data, data, tsunami, data, glut the other side there is a huge competitive intensity which is changing the consumer saying that look be more relevant for me and that's the context in which really analytics is playing and therefore my view is that we're really talking about analytics in the context of saying how do you take improved decisions what is sexy about analytics is the decision making capacity and really that's what we're trying to talk about as the capacity for analytics and when you look at this chart in most companies the x-axis you have a lot of low volume decisions which the company might make someone walks in maybe once in three months he has to be given or sold a product that's a low volume decision but someone calls up every day asking for something that's a high volume point of interaction and the other side you have the value of that interaction so in the high value very low volume kind of situation could really be like a merger and acquisition situation where again you can use analytics but in the case of consumer businesses you have much more of these high volume and lower value interactions which happen with the business everyday transactions and how do you bring intelligence into that is absolutely key the ability for businesses to go back and work on the technology stacks connect up the big data and connect that with rule engines and ability of powering the right algorithms to take the right decisions then becomes a huge, huge opportunity so really what I'm saying is that effectively analytics needs to start by listening to the consumer because if you are as a consumer you don't want to be treated like a market segment you know you are not happy unless the marketer actually speaks to you as you right and how do you make that happen and how do you anticipate needs without doing that you cannot expect that loyalty to happen so that's the context within which you are operating and if you don't do it the consumer goes so it is really not a choice that businesses have today because you are collecting so much data let's start to use it the context within which I am talking about is that there really is not too much choice that if you don't use that data better and actually look at the footprints which your consumer is leaving behind you are going to get left behind as a business and to do that most businesses and all of us who experience this as consumers tend to you know look at the product and say here is one offer which I want to sell to all of you right and so it's a mass rendition of an offer how do you move from that paradigm to a paradigm where you have knowledge because of your data of a person with a change in behavior suddenly this customer who was never coming in earlier has come into my store today that's a change in behavior and now how do I populate the right offer the most relevant offer connected to the change in behavior and that's the context in which analytics starts to work and that's the context in which you can actually go back and justify the ROI around using all of the technology required to distill insight out of that big data so if companies have to do that and if they have to really do that effectively the biggest road block is what I call the silo elephant right so we are talking about the fifth elephant here in the conference what is a silo elephant the silo elephant is the problem that if you want to improve or change the way decisions are made within companies you have to break silos you have to be able to go and speak to the CFO and tell him why you're doing something in a certain way which is different in the past you have to speak to if you're a retailer to the merchandising side if you're a bank you have to speak to your credit risk people so at the end of the day however good you might be in the crunching of your data if you don't start to tell stories what happens is that you don't influence the decision makers so even before you can influence the kind of decisions you take for consumers you have to influence people within your company and break the silos and if you aren't able to do that you aren't able to influence and the proposition very clearly therefore is that if you are an analyst sitting in the context of a big company or even from outside as a consultant you have to find ways to be able to tell stories to be able to make people change their mind now what could this be about this could be about for example if I'm a bank and I'm selling loans I would go and actually market my loans to everybody but I have an internal base of customers to whom I might be able to sell loans far more profitably but then for that the sales channel has to think differently I have to assemble data make it rational enough for people to buy the argument and that's the whole aspect of storytelling and what's happening more and more is that India is a country for example with more than 60% of the people being less than 27 years of age so you have a younger and younger consumer profile and in the business world therefore younger and younger people are obviously taking more and more distant making positions and what they are getting used to is consuming data in many ways I think all of us today have so many apps which throw data back at us and they are getting more comfortable with that that people younger are actually far more comfortable with data than people older but the changes happening storytelling with data will start to become even more critical and probably more naturally done by the younger people and will be critical from the point of view of influencing change so I wanted to quickly take you through an example of a retail data set and very briefly tell you a little about that I know that on a small size over here I'll speak a little more about it but before I go into that how many of you have heard of Hans Rosling how many of you have seen the 200 country presentation so as I haven't got that many hands up what I'm going to do is just switch a little bit sorry I'm just switching this a little bit and before I start just a background so they say that if you have to tell stories with data you need to have PhDs with personality so how many PhDs with personality are there in this room I'm not a PhD so I don't want to make any judgment on PhDs with or without a personality but the challenge always is that how do you get people to tell stories really really well and I wanted to show this video to you to tell you about a guy who I think is probably the world's best in storytelling sorry for the technical edge yeah so while we're setting that up so what happens when you're doing this is that remember these stories that we're talking about are to be told on data which people are very familiar with so obviously what I show you will be will be very sexy because it's completely consumer knowledge kind of information but when you're doing this with business data obviously the way you do it will be different right but the basic notion I want to leave behind with you is that unless you unless you do that you don't make that impact to be able to get people to start to swing their thinking to take this in a different way so here we go first an axis for health life expectancy 25 years to 75 years and down here an axis for wealth income per person 404,000 40,000 so down here in 4 and 6 and up here is rich and healthy now I'm going to show you the world 200 years ago that's it here come all the countries Europe, France, Asia, reds, Greece, Queens, Africa South of Sahara view and the American yellow and the signs of the country bubble show the size of the population and in 1810 it was pretty crowded down there wasn't it? 4 countries were taken for life expectancy will be very important in 1 country and only in the UK now why stop the world no? sorry I'll switch back sorry I'll just switch that's okay we'll go without it okay I'm sorry you couldn't you could hear much of it but at least the point that I wanted to make for all of you was that that you really have someone out there who can tell stories very effectively and if you're able to leverage that skill within your own companies and get that kind of skill sets in your own teams that will be very, very useful so I just want to take you through a little bit of background here is an example of a particular store or a particular retail business that we've done some work with I've masked the data but you know the challenge that you normally have in retail is the fact that retail is a very catchment led business so consumers come in from neighboring catchments so if you're living within 5 to 8 kilometers of a store you would come in and use that store so in this case for example we took 3 years of data and we tried to put that in to try and see what sense can we make from it what kind of business decisions can we make from it so I just want to quickly take you through some of that so in this case for example we tried to map the data and we said that on the y-axis we have the number of loyalty members that business had and on the x-axis we saw there the performance across month and then how did these customers do per each month going in from 2009 so when we did that when we did that what we tried to look at was that month on month is there a pin code which is kind of really really differentiated are there some pin codes which are screaming out more than others and you know appropriately one particular pin code which you can see right on the top which is that it not only got more members it also got much more sales started jumping out and when we started to see that when we started to see that we said okay what is that pin code and we focused in on that pin code and as expected because intuitively retail business does come from neighboring areas there is a pin code around the store so when we did that when we looked at that pin code we said okay so this is a pin code which seems to be taking the lead again how good are you as a retailer in knowing that pin code do you really have information about that pin code do you really know what your store members are doing how much they are really using your store are they shopping the full width and things like that so when we started to do that we looked at this pin code much more in depth and we said how is this pin code doing so when we started looking at that we found that there seems to be a steady increase in business so there is a continuous increase till around September and then suddenly from that point there was some stability and there was a spike so we were able to focus on two months where they seem to have been in September and October a sudden spike in business and the amount of business done per customer seems to have suddenly changed so we went back to the store with this three years of data and we said that if you've looked at this this whole movement of data and if you really looked at this movement carefully what you find to take you through that in a summary is that a lot of people came from the target from the target 40064 which was the pin code which jumped out but a large number of people came from 10 km away which in terms of the Mumbai map which 10 km away was actually 1 hour 15 minutes away in time so what this meant was that the retailer could then think about location strategy in many more interesting ways so the reason I am giving this because the only point I want you to make with this data is to say that now you could start getting even more interesting facts together you could pick up locations of all competing stores, populate them look at data which you might have from other external sources bring that in and we did all of that so while I don't have all of that story what we were able to do as we assembled this data is that when we got into a room with their distance maker who wanted to take distance around where to have those stores we were compelling we were able to tell him that look here are these three areas which make much more sense than others so I think the point I want to really make was that you really can make a huge difference when you look at data find insights but focus a large amount of your time in distilling the value or the storytelling value which can help take distance so that's the point which I really want to talk about I have gone through this like an express train but I will continue the same pace I think if there are any questions in between I am happy to take them right now we can take questions while I am speaking I am okay so if there are any questions I am happy to take them right now anyone yep so if you remember what I said you just repeat the question the question was that what do you mean by data equity so you have data which is there in the company and the data is about the customer and let's say it's a bank and I am the customer and let's say you are a customer and let's say you are Bill Gates so I am me and you are Bill Gates and we both are saving the account and if 300,000 rupees is coming to my account what do you think would that be significant it would be right so for me it is significant if 300,000 rupees is coming to my account it is significant but remember you are Bill Gates if 300,000 rupees is coming to your account is that significant no so now the bank has the data to say that the same 300,000 rupees which are for me are significant should now have a distance which say that some relation manager calls me saying would you like to invest insurance which is my data and your data has allowed the bank to make an effective communication to me and tell me to do something which is very specific and which make the difference in my life or if I walked into a retail store and when I am in the store while I am buying at the checkout point if I was told that you are buying so and so but because you always buy so and so here is a special offer for you again that's bringing the data right at the front end so as you start to see this from a company repeatedly you start to see equity you start to say hey look this this makes sense to me and you start slotting that company as a company which relates to you which understands you finally what are we doing through brand building by building brand equity that's what you tend to do that's what I mean by data equity being more powerful any other questions when you are doing prediction how much weightage do you give for your historical data for your historical data historical data to predict your sales how much weightage you give or what all metrics do you consider so if it's a forecasting problem like predicting sales then you need to cover much longer you know time periods because you will also have seasonality and so on because you might have a business which is seasonal so you might have to cover certain amount of seasonality elements for you to be able to forecast so you would look at historical data going back to at least maybe at least 2 to 3 years to be able to make a prediction but if you are looking at predicting what this person might buy you can work with very recent data and fairly decent predictions can happen with fairly recent data and that can allow you to take this in as well I will repeat the question if you just say it can you hear me can you hear me sort of my question is when you are faced with a large volume of data and you want to tell a story out of it it's such an open ended problem and it's really you can do tons of analysis on any set of data so what's the process that you use to actually look for the story so the process essentially is what we call POV so essentially as an analyst you need to have a point of view so let's say this big data that you have is about let's say a business where the business is not doing well you need to have a lot of hypothesis that you which you allow to be tested and if you don't go in if you expect the data to talk to you without that it doesn't happen you need to have very very aggressive points of view and that is where again I will repeat that the intersection skill sets are very very important because people when they come with different backgrounds you know make a huge impact in terms of being able to produce those points of view and those points of view are very intuitive points of view not basis data and the story comes out so that's the first thing that you would always do have a very very aggressive point of view I don't know if this is a good question to ask in this context but if I were to think of this in purely in the case of an online scenario how would you deal with noise because the example that you gave the amount of noise there is relatively very low you're dealing with transactions that have happened but in the case of say a purely online business the amount of noise is very high so how would you filter out noise see I think while the noise is very high you also as against retail store online retailer can very very constantly experiment so if you are really I think possibly the paradigm of how you do analytics will be very different and you might have to go in with a lot more experimentation and take the results of that experimentation and yes there will be noise and there are techniques to be able to help you with trying to sort the noise problem out but I think by continuously experimenting with what the consumer is doing and picking up and building on the experimentation data and then coming out with the next step is the only way to do it in situations which are online I'm sure you're doing that kind of stuff but that would be the way I would approach it so test and control design of experiments doing a kind of a factorial design setup and then making sure you create a lot of options for consumers to look at and see what works and what you see is through data so build a protocol for testing which is scientifically built and let that data talk to you more and engage you far more anyone else what do you think that once I've told the story about the data so what is the success criteria for the story so I think the success criteria for a story is that you have been able to take the decision that an analytics told you is the right decision so if you said that you know that if you looked at replenishment in a retail store and you felt that you need to change the way you replenish certain stock and you did some analytics to be able to figure out what should be done in the supply chain now your story needs to therefore connect to the decision and therefore it's not necessarily only one story it's probably a sequence of stories because there are many stakeholders you need to tell stories too so I was using a metaphor of story just to say that look there are conversations you have with multiple decision influences with which you finally but finally the decision which analytics tells you you must take if that is not happening you may have a great story with lovely insight and no decision that's like operation successful patient how do you decide like how do you decide what should be the granularity of data you know when you're modeling sometimes we can tend to over parameterize the things you know when you're modeling a real life problem what is the essentially how do you decide these parameters matter this doesn't and you can over parameterize when you say over parameterize you mean width of very deep data or so lots of data so basically in a retail store example we tend to connect with let's say is there any correlation between weather and the footfall should that come into the picture or not so I think as a general principle obviously the wider the data that is that you have is good but I think it's my belief that if your data is not setting on hypothesis it tends to lose business context because again when it comes to you might say weather is very indicative weather may have other dimensions that play into you know into the data and for example traffic may get worse with weather and the noise starts to increase so I think intuitively you have to look at the width of data and say that will I be able to take a decision based that kind of data and actually ruthlessly take out the data which is not supportable from the point of view of decision making you know that's one second I think the more the data within the hypothesis that you have the better the deeper that it is the better but I think you need to figure out how recent is your data and if the recency factor which matters matters much more so I would look at customer behaviors changes and whatever data helps me spot that changes that's good for me and I don't get locked into too long history longitudinally that's the way I would look there are many incidentally points of view on this and multiple people have different ways that's the way I tend to look at it just one last question I am on statistician actually and I have a lot of problem about how to I can do the analysis maybe but I am not able to tell the story because I am more obsessed or I am more thinking about the assumptions and all that that get into the modeling so how do I present my story with the underlying assumptions that I made for analyzing the data yeah so I think as a statistician you are you know obviously much more focused on the technique and the stats behind it but I think there are a lot of insights that our statistics is taking out and if you try to say that I mean one good example which I use and I have frankly failed on that a lot but I have a 15 year old daughter and every so often when I have an important presentation I run a few slides through with her and she says that's garbage I didn't understand a word of what you have said and so to be intuitive common parenthetical words and intuitive business explanations are far more important than saying that you know that case factor is not right or you know the business guy doesn't understand it so if you want to say case factor is a problem can you give an example of separation in a room which has some people who are black and some people who are white you know so can you try to work at trying to find business oriented people intuitive explanations for a lot of statistical phenomena and then there are a lot of stuff on the net which will help you on that you should look at that but simplification and business oriented language is what helps just want to wrap up with just a few thoughts so one basically with this whole data led marketing interesting conclusion that Gartner has is that by 2017 they say that chief marketing officers will spend more on IT than CIOs so to that extent obviously with digital with web with consumers leaving so much data there is going to be a movement towards marketing consuming that much more data and obviously as more and more consumers start to engage with businesses on mobile and on the internet you start getting a lot of pull behavior which is basically that I am trying to download something so that's a lot of interaction behavior and as you start to understand that interaction behavior it gives you a lot of information about how customers are changing and I give you the examples about the jewelry purchase in a credit card situation I call those events and those are really event triggered examples and there are lot of such events the Bill Gates example was also an event so how do you powerfully use customer behavior changes to spot events and then drive businesses around that and therefore link all of the work that you guys do in analytics link it to the last mile and see what impact you can make in that and finally I want to leave you with the thought that analytics the reason why storytelling becomes so important is that analytics finally has to solve a social problem it's not solving a technical problem the logistic regression methodology algorithm was done 45 years ago we still use it for cross sell we are solving social problems and if you go back and look at the intersection skills that's required to build this solution using storytelling I think there is a long way this can take you thank you