 So many of you just you know Kevin told me okay. Hi Jenny So I believe you all know me or you know if you if you don't so probably you can just go and see in LinkedIn There's no point giving an introduction seriously so I'm into media industry and I work with vacuum 18 and as a if you don't know vacuum 18 But I believe you will be knowing MTV and you must be knowing big boss. So those are our productions Currently we are running Spitzbilla sunny Leonie. So that's all our productions and I mean, that's how you should associate vacuum 18 with we also have a digital arm called vote and Many of the analytics, you know the reason why Suddenly media industry required to do analytics because they all of a sudden when the digital arm of those media industry started They started getting a enormous amount of data all of a sudden before that They only had bark data, which is based on 20,000 to 30,000 samples and beyond that They cannot even assume that there could have been any data So all of a sudden when they got that, you know every day couple of million people coming and you know browsing their app Watching content. They suddenly found. Oh, wow. What do I do with this? And that's exactly where media industry particularly in India started investing in data science And it's a very good junction. Even if you know, many of you guys are in different industries I must say media currently is media sports and probably health care or some of those industries Where data science is going to grow? Really in a very different way that is the emerging industry and I think this is the best time to be in the media industry And in data science. So I just wanted to give you, you know few games And whenever I speak to people the first question they ask how do you connect science with entertainment? Like ideally we're creating some product some sessions some episodes and you know some movies How are people going to you know use science into it now basic science like, you know Customers analytics revenue analytics those are there those you will find in any industry you go to retail e-commerce Banking telecom any industry you go you will have those things the biggest difference in media industry That you don't get anywhere else is the content analytics So if you are a person who loves to work on the new these toolkits like speech recognition video breakdown or the video analytics You know like of face recognition those kind of stuff sticks analytics all those kind of stuff you will be able to do in the media industry and That's where the beauty of media industry this kind of work is almost not possible in any other industries Sports is of course coming up and that's another place where it is You know very different use of analytics and data science will happen But media is a very very very new one And this is this is actually a joke before we start the use of science This is the abuse of science that has happened So how many of you remember a movie called Ravana that was made in Telugu first and then it was Remade in Hindi and there was a sequence where there was a song that And then someone thought that they can actually translate run a subtitle of this song Right, and can you tell me the craziest subtitle that happened out of this? Yes So it actually came out to me because the way They realized Bane is connected to sister. They is give so they translated it give sister give me your sister So yes, I mean if you don't do science good enough and if you think the Basic translation or transcriptions, you know work can happen This is the kind of you know challenges that you are going to face and this is running on national TVs with this kind of Subtitle and actually going to international Imagine what people will understand from this song or about the movie So as I told like you know earlier It was a more big screen media today It has all came to our pocket and more it has came to our pocket the data actually increased and it has increased Exponentially it didn't increase in a linear fashion So data science in media industry almost touches every nooks and corners of this business I'll just give you few gleams in different, you know six Sections where it does as I told you marketing, you know It's mostly customer related analytics that you do anywhere else consumer lifetime value acquisition all those types On the ad sales is actually something very similar to who who of you work in a supply chain kind of Domain like how do you design the right route map for a olacab is something you can actually go back and apply in the Media industry also in terms of when you are selling the ad spots So what what we sell actually in media and what earns us money the way we sell our ad positions So how many ad I should put at the starting of the program? How many mid-roll ads like in the middle of the program I should put How do I even design the pricing around that? How do I distribute and sell the consumer profile all those things require a huge amount of Optimization works that is you know required to do which is very similar to supply chain kind of work So if you know you people are working on supply chain, that is a very ad ops ad Optimization is a very big domain in the media industry The biggest work is happening around the content creation and content curation and that is exactly where I'll give you some Examples and the use cases and those are real-life use cases that we have done in Wacom I think So what we are doing is we are working. We are building AI models, which will help us determine How to green light a content? So what is a green lighting? So there's this is a very no typical terminology in media industry. They say content lighting In most media organizations like you go to stars the Sony Wacom I think anywhere There will be a bunch of people whose job is every day just sit and watch the different movies different contents Listen to the different storytelling by the writers looking through the screen-written things all the stuff That's what their job every day They are paid to watch movies because they watch those or listen to those stories and come back and decide By their judgment whether we should produce those stories convert those into a full-fledged movie or a Series or a wave series or whatever we want to do now that has an hundred percent human bias And that is the reason why they believe this movie will do good But it doesn't right on the other hand at times They believe that this movie may not do as good, but it time, you know, it will just do a box office hit So we are trying to create a AI driven model which will ideally Given a very detailed screenplay written will tell us what is the probability of the success of that if we convert into it That will also tell me therefore how much I should You know use for the production cost of that particular piece and how much I should not spend beyond which my profitability Will not be there therefore who are the cast I should choose what are the prices I should pay to those cast if I choose cast a versus, you know actor a versus actor b What is the success rate changes? So all these kind of things we are trying to build We are also so whom to cast how do I design the dialogue? We are even working with the narratives who are writing all the detailed dialogue. So the way, you know Our story is completely full-fledged done first comes a simple narrative like a basic story From that someone writes a screenplay so screenplay means it decide, okay These are the different scenes this scene will be shot in this kind of section. These are the different people These are the basic story then moves it goes to a dialogue writer who actually fits in between and write every Person's dialogue inside and then it goes to the movie producer movie directors or whoever are producing and directing those To ideally decide the final thing what happens is when the screenwriters are writing it versus when the Actual cinematographer or the videographer is shooting it They change many things on the spot because what is written and assumed is not looking good on the spot So they change it so ideally what has been thought through versus what actually comes out You will find close to 40 to 50 percent difference between them We are actually trying to help this people that what kind of dialogue fitment is going to help you How do you do the talent management? For example very recently you have found that Netflix has been trolled because of Radhika up there that all the stories That's coming around are of Radhika up there now I'm sure they have done some kind of AI where they found that I mean most of the movies or you know web series that they do of Radhika up there has picked up very well particularly they did a series of Rabindranath Tagore story by Radhika up there where most of those stories she was the You know lead protagonist and therefore they come up with that We can actually go and sign a contract with her for a long time and we can produce many many series now They can absolutely go and do that But how do you publish do you publish immediately one after another like after sometime people will say no More Radhika up there now that's not good. So you can produce those but probably Time timely publish them so that you know you have multiple I mean many other stories and not just too much of Radhika up there It's like if I continuously keep you feeding chocolate after sometime We will stop loving chocolate right so it's the same thing So how do you actually choose and do the right talent management there what kind of licensing and contracting we should do Content editing so this is another place. We are actually working on artificial intelligence and So suppose, you know, we have big boss now in the big boss house. There are 50 cameras Continuously following all the participants there all the contestants there what you see on the TV is only one hour or one And half hour content now There are 50 cameras has so many different other contents that already there which are ideally thrown off Can I actually create different story lines which are like back of the scene which are not shown on the TV? And how do I create that? Based on what we understand from what has been already shown on the TV or what has already been published What section of the stories people are watching re-watching or not watching and based on that Ideally strategize what should be those, you know other cuts that we should create or suppose we have a Say 10 hours long content. I want to produce a very short highlight of that Which will be packed within 45 minutes. How can I create that? We are even experimenting now to give the Gemify all this thing where we will give the power to the hand of audience We will it's like, you know, how do you cook a recipe? You have all the ingredients, you know, which muscle are to be put, you know Which all how it should be put and then the final product comes is the exactly same way If I tell you these are the protagonists. These are the different emotions you can mix These are the different actions you can mix you choose and you create your own highlight on the spot and Artificial intelligence will be able to cut those kind of things for you and create your own specific highlight So where it will no more remain big boss what I want to show you you craft your own big boss, right? So these are I mean I won't say that all of them are hundred percent done now Many of them are as a POC pilot those stages, but we have found a huge success while doing this Content licensing it's a small work where we talk about when we are licensing content and not really Creating those contents ourselves are should we do that like you know buy a perpetual license? Or do I actually buy a short-term license and then sell it back to somebody else? Content publishing is purely based on the recommendation engine and you know, how do you publish when do you publish to whom do you publish? Customer satisfaction, I won't go much But product design is again another thing very important for any app based Technologies whether it's a food take or it's an e-commerce or anything because Easiness of navigating that app is always helps consumer stickiness and help consumer to come back to that app again And again, so how do we make sure and understand? What are the struggle points that consumers are getting through and how can I keep on removing those struggles and make that app? Lighter easier and you know people will start loving it. So that's where the product analytics is product design is very helpful. I Will quickly you know take you through a couple of things we have done So generally so far food analytics has sorry food has always produced web series Which is six to nine episode long not really 40 episodes like Netflix So what we tried to check is how much people binge watch do people at all binge watch? And when I release this series Is it good to release all the episodes at a time or is it good that by story? I break it down and people can just come and you know snack those content over the time And then I keep on publishing probably every week in a certain interval What we have found is people love to binge watch in India But there are different categories of people and how do they binge watch? So a 40% people 4 out of 10 binge watch the whole series at a time And you know this people mostly watch it by the midnight like that their clock starts around 11 and ends by around 1 a So on our consumption pattern we find a huge jump at that time So there's a good bunch of consumers the moment is it is published They just start complete it at one go and then close it There is another bigger bunch who actually watch it in a one hour long period So don't watch together at one sitting, but they will you know break it break it They mostly watch by the afternoon time. These are the people you know either they are shopkeepers and What we thought they what we first hypothesize this might be the people who are housewives and by the afternoon time They have lots of time when they can watch it We were surprised that 80% of these people are male and we found that okay. How is that possible? So then we found these are the people mostly who are the shopkeepers or taxi drivers or those people Who do not have much work during the afternoon time and when they're actually watching it even when they're sitting in their shop They don't have much customer. So they are actually sitting and watching at that time and therefore their works comes an interval They keep on watching you know as they are finding time So therefore they don't watch at a time that three hours kind of thing But they watch over one day. They close it. They complete it Then there is a bunch of people that generally take one to two days And these are the people when they take one to two days are weekend watchers So even if I publish something by Wednesday, they don't come and watch immediately a search comes during the weekend When they go back check what all new thing has come during the week They will only watch during the weekends and they will close it the last once they take longer time more than two days And these are the people who are mostly commuters So suppose you are in a train or metro or in bus you download this contents and watch as you go And these are the people therefore take more longer time because they only have those commute time and therefore They're using those time to watch they take longer time for some of them Yes, because otherwise we had no other way to understand who they are and if we just would have you know Thought our hypothesis right we would have stuck that these people are Housewives the second group but when we found their mail then we thought that okay What is happening like who are these people because it's not possible in office during lunch hours You are sitting and watching you know some content is generally not possible in office So then we did some survey in the bigger cities the smaller towers to towns and there we found Yes, these are the people who are doing it and then the hypothesis that when we should publish So we found a sure sort answer that we should publish all the series at a time The biggest reason is otherwise the continuity of watching the content breaks and It takes a lot more marketing money to bring those people again back next week to the same story because People buy that time somebody else might have launched a new show and people already move out of the older show So it's better to launch them and also the dotted line is the completion rate So what percentage of the content gets completed if I launch them at a time? It's like really shoots up to 1995 percent and stays there if I launch them all at a time because of the binge watch If I will launch them bar. I mean in our in our blocks. It doesn't happen. It actually has a break Now this is an interesting story where we found that most of the originals that we launch from first to second There always is a drop but those who already come and watch the second episodes They will almost you know 85 to 90 percent complete the whole series for all of a sudden one of the original We found a huge decline first episode was getting a record number of people and Surprisingly second episode the conversion was very very bad and it was drastically bad So the question was what do we do should we market the first point always is okay Hello, go and send notification come and see the second one There has to be some natural conversion from episode one to episode two and we said there has to be something wrong in the episode One why people are dropping we have to figure out their drop-off point So we went and analyzed people every session level data that where they started from the content How much they watched where they paused where they scrubbed so scrubbing means you are first forwarding the scene From which part they are watching and beyond which they are not watching now either if you are dropping off Then I know that around that point you are not liking the content or if you are scrubbing I know that portion is becoming boring therefore you are not watching so when we analyze those details We clearly found and we plotted them as a hit map. We clearly found there are three sections Where the biggest boredom is happening on that first episode and You know almost all the eighty percent people who are dropping are dropping around those places We went back and watch the content exactly at those periods and we realized why it is happening We watched it along with the content people who are creating those content or editing the content So we said go and read it So they re-edited the first episode relaunched the whole episode once again with the re-edited first episode and we found almost a 23% uplift in the conversion compared to where it was dropping Just to keep in time we are doing many more things in terms of we are breaking down the videos now Microsoft Google or Amazon has good APIs And there are other more intelligent APIs through which we can break down the videos and create lots of meta tags or AI Generated meta tags which otherwise not possible For example, who are the protagonist playing at that level? What are the emotions of each of the protagonists? What is the kind of scene? Is it a bedroom scene? Is it a outdoor scene? Is it a you know sitting room or a kitchen scene those kind of stuffs? We are also checking that you know, what are the Storytelling around that all those things putting together. What is the color palette that has been used there? The protagonists who are present what percentage of the whole pixel size that they are actually present Is it a back face? Is it a side face? Is it a front face all those kind of stuffs and you know Connecting that with the viewership that we get at every nanosecond viewership We are connecting with that and getting some very interesting, you know Understanding that what is happening? Why some percent some people are watching certain section why some people are not watching? How can I reduce if in an episode there are lots of certain kind of emotions? Which people are not enjoying or a very flat kind of episode? Chances are very low that people will come back for the next episode. So how can I remove those things? Ideally helping the narrator or a screenwriter or a dialogue writer to change those kind of sequences So this is actually from one of our regional content. This is a long Series where we found that so these are the same person this is the same character and If there are more negative situation like you know the person who is against her and literally Hitting her making her cry the viewership somehow increases up pretty significantly whereas if she is Happy and you know nothing happening people are like actually nothing happening Then we found that you know when certain Protagonist is coming for a longer duration people are switching people are switching from that either on the TV And this is where we are actually mixing the digital data on top of the TV information So this is something a very big breakthrough We have worked on to establish a very common currency between digital and TV about the measurement And this was a very tough job But we still have done it and we can actually now see on the TV Because TV actually you know that when people have moved after that one minute even though the granularity is one minute But you still know it and on the digital also now we can exactly map those and we can see where people are dropping off or switching their observations Yes, yes So then we have also found that you know What is the collaboration between the hero and the heroine of that and you know what kind of emotions or situations are working when? Both of them are probably in a very simple situation going shopping or other things people are not so interested When they're talking about you know see the daughter-in-law talking about the mother-in-law to the husband And you know people suddenly start checking that okay, what's happening? And you know if there is a negative sentiments the best part we have found is there are lots of Apart from the main hero and heroine there are lots of peripheral artists also in that now those peripheral artists are actually playing a very Important role in that story, but if they are only alone it doesn't make it doesn't give a uplift But in combination of those with some other person gives actually a bigger uplift in terms of viewership In terms of stickiness and also certain kind of emotions So what we are doing is we are checking at a first level is about the cast the combination of multiple cast combination of cast along with their emotion and sentiment and the fourth level combination of cast along with the Location of the short plus the story and the sentiment and the emotion of the people all these things put together What's the what's the actual recipe that has cooked and what is the uplift that we can see and that's how we are helping our content curation Team to actually curate a better content and this is what we are doing not as a Post facto this is when actually the series is already running every day 30 minutes It comes so as we are learning we are continuously giving the feedback and in our you know Industry mostly the shoot happens like when it is airing on TV probably the shoot has completed just two days back So you really have good amount of time to go back and give the feedback That okay You should need to change the story lining on this side the story that you have started people are not picking it up on That probably you need to change you will find lots of sudden change You know all of a sudden a character dies or something happens because they found that character is actually not working well for them Or actually the face change happens, right? So for like for example Balikabad who when happened there was a face change that happened, right? So lots of these things are happening based on this analytics behind so thank you so much I think I have overdone the time Okay, I'm so sorry. So any questions I am here you can ask should we take any question now or later? Yeah, so today I'm around I'm here in this hotel you can just call me what's up me like me LinkedIn I'll be always happy to answer