 Today we have a very young passionate data science enthusiast who is a leader community member and what more so she has hosted many meetups participated in many of the meetups community events and She has a lot more to tell you. So maybe I'll invite her for the day. So without any further ado, I'll just hand it over to our next wonderful speaker. Rajneeth Kaur. So over to you, Rajneeth. Much. Hello everyone. We're going to have lots of fun together on inside generation today. And let me tell you a little bit about myself first. So I'm Rajneeth and I've been in the space of data science for over almost a decade now. I'm currently working as a senior technical project manager with persistent systems. Just an hour back, interestingly, I became part of the top 10 women in tech award for her rising and I've also been shortlisted for the women in IT Asia women leaders series. So, in terms of my 40 insights generation time series forecasting promotional analytics NLP and ML in general is what I would say. I am passionate as well as good at right and at any point I'm handling about 25 to 30 data professionals. Right. That's about me. Now, let's get to you. The reason that you are here. So this session is going to be for you regardless of whether you are a student, whether you are an existing data analytics professional, you may be a business professional like a marketing manager or operations manager trying to take better decisions through analytics. Or finally you may be an existing analytics leader. I promise we have lots of interesting findings and takeaways for each one of you. Now, one thing that I keep telling the hundred plus people who've ever reported into me, I'm going to be sharing that very soon. But what we're going to start with is what is an insight, the process of it, and then we're going to apply that through case studies. Right. So this is what I've told everyone who's ever reported into me that data science is as much an art as a science. Right. And what I want to do hopefully today is break some of that art to scientific chunks that you can take away with yourself and implement right from tomorrow. Now, what you're going to really ask me first is what really is an insight. It's a buzzword that we speak all the time and really leaders, clients, stakeholders today want us to give them hard hitting insights. And lack of insight will lead to lack of engagement and possibly loss of business as well. So what is really it? Let's start with what it is not. And I'll let you enjoy this little comic where you'll see big words like actionable analytics, in-memory computing. So the first thing I want to tell you is insights are not data, they're not technology, nothing to just do with business and any jargon surrounding the same. Insights are not even observation. Right. The way I like to remember what an insight is through this stakeholder telling me that I'm trying to make a decision, stop confusing me with facts. And if there is just one thing that you take away from this session today, I would like to tell you is remember an insight is something that enables better decision. It may be a composite of many different findings. It may be a hidden pattern, but end of the day, this has to be sticking around in your mind. We are going to delve a lot more deeper into the generation process, but generally what you would get is a big problem statement like how am I performing, right? And you have to break that business objective or primary objective into secondary objectives or analytical objectives that can be answered through data. For example, how am I performing can be broken down into is my market share consistent or rising? What's my year over year change? Hopefully that's a positive number. What's my attainment percentage? Which means that if my target was 100 units, how many did I actually sell? Let's say if I did 90, my attainment would be 90%. If I did 120, my attainment would be 120%. So if all of these positive numbers are there, then you've definitely performed well, but we're going to get into much more depth here. Next, I would like to say one tip that I would like to give from a storytelling perspective is that cater to different types of learning styles. You'll see in your environment that there are people who learn better, who look at a graph and they understand in a Jiffy what's going on. Another person may be a numbers person and a third person may like you to give a crisp English line, right? So used mix of tables and graphs and commentaries to do your storytelling of an insight. Now, this is what I would like to ask you, would you like to have a checklist? I think very few things in life are comparable to the pleasure of knocking off things off your list, right? And very often we get confused between what a finding or observation is and what an insight is. So my guess is that you do want to see, so let me take you through. An insight is something that is big, useful and surprising. So we are really talking first of all about a large number, a large impact. Second thing is something should be doable about it. It has to be actionable. Finally, it has to be something that's new knowledge and not very obvious, right? Let's take a couple of examples. Apex region wise share increased by 1.3%. Now is 1.3% a big number? In this context, probably not. Can you do anything about it, my dear friends? Can you action anything at a region level? Or if this was at maybe at a city level or a store level, you can change things. But at that high level of granularity, not really. When it comes to it being non-obvious, trust me, there are a ton of BI reports that your stakeholders as well as clients would have seen long before you get to these facts, right? So this is not an insight. Let's look at another example. The sale of crackers increases during Diwali by 1.58%. Now it's a huge number, 1.58%. But is it actionable? Can you change the date of Diwali from November to October, for example? Is it surprising even? Did you already not know the correlation between sale of crackers and Diwali? So no, again not an insight. Let's take a third scenario. In this case, let's say in a recent month, your sale has piped by 120%, which means that if you were selling 100 units, now you're selling 220 units. And you want to find the root cause of it because you can repeat this success story, right? So you find that you recently started investing in LinkedIn and because of that, the spike happened. LinkedIn is giving very good ROI of 2.3. So all big numbers, first of all, large impact. It's very useful because now you can start investing more into LinkedIn. And it's surprising because it's a new event, right? So a very simple and elegant framework by S Anand, the CEO of Grammler. But I would like to call out one thing here. It's not as easy to implement. Especially that surprising bit that you'll see is very tricky. And very often I've seen that something that's surprising to me is not surprising to my stakeholders. So what I would suggest is over a due course of few months, try to understand the daily lives, the challenges, KRA goals of your stakeholders and clients better, and you'll be able to understand what's surprising to them. And on the other hand, you can also make them accountable for telling you what's surprising. So that is about the checklist. Now what you're probably going to ask me next is how do I even come to observations like this? Because trust me, if there are 10 observations you come up with, maybe one of them will be an insight, right? So how do you get to this observation part as well? What is the process and what's the approach to this problem? And one thing that I would like to again share that's worked beautifully for my career have been business frameworks. So there are a plethora of them that exist and they can help structure your problems, analytical problems much better, right? I still remember during my MBA at one point, my marketing professor said start with a framework, right? And everything else will start falling in better. Let me share a few examples, right? So product price plays in promotion, 4P. Anything descriptive that you want to do in terms of marketing problems, this framework comes very handy. The second one Pareto, the AT20 rule. Does that ring a bell, at least for a few or few? So this is the law of vital few and it's very good when you want to prioritize and you want to pick few things because of which there can be a major change. And some examples are 80% of the work is done by 20% of the people or 80% of riches are held by 20% of the people. So that's how it works. Now it doesn't have to be AT20. It can be 1 and 99. It doesn't even have to add to 100%. It can be 2 and 57% for example. But the law is for prioritization. Business is also full of 2 by 2 matrices or quadrants like the BCG matrix. And stakeholders, trust me, love it because you have 4 boxes which is enough decisions to take but not too many. So you can target each of those quadrants differently. There are a lot of analytical frameworks as well like hypothesis testing, 5 by where you keep deep diving into something till you find the root cause. Cause benefit works very well when you're trying to evaluate goodness of a decision. MECE and then RFC, Recency Frequency Monitor is very famous in the marketing world, details especially. Now this slide has become very heavy. So I'm going to, we'll be sharing a link with you about an article that I've written about using business frameworks for better analytical solving. And I would recommend research and read upon them because they really aid the process well. Because I would like to use the rest of the time to take you through some case studies. How do we actually apply them and help our stakeholders make better decisions? So we'll be sharing that link in a jiffy. Now one thing that keeps happening across client calls, right? So we have, let me introduce to you Mahesh and Karli, the heroes of this show. Mahesh is presenting to his client and he states that, okay, the reasons could be that EMEA has fallen 20% year over year. Product Pandorab that's our new launch has not been meeting our expectation. And then he goes on to say something about offline channel but Karli cuts him short. She says that don't tell me all of these facts. My attainment is low at 90%. I could only meet 90% of my target. Tell me what can I do about it? Can I explain why the target was unreasonable? Can I ensure in some way that the next time I'm not in the same place. So what better decisions can I take? So what Karli is really expecting is an insightful discussion, right? So if I were in the place of Mahesh, what I would start with is some descriptive framework like the 4p and have very much detailed questions around it. And everything that I'm going to show you henceforth guys, I have implemented at some point so they're totally doable, right? So for example if it is product. The kind of questions we could ask is has the product mix changed? Let's say I was selling A and B as products. Maybe I was selling them equally earlier but now A has started selling more. Or maybe there are some new launches whether my own or a competitor launch. Another thing that could happen could be changes in pricing. So the first thing we could try is take the trend of price and plot it in a simple chart and see how the average is trending. There could be discounts as well. Maybe earlier we were having discounts. Now we are not or vice versa that could be leading to no attainment. When it comes to the place aspect, there could be something interesting at geography level, whether at region, country, city or school level. Maybe my online versus offline channel mix has changed. Maybe I'm investing more in one of the channels than I should, right? When it comes to the online channels, one thing that could be happening is that between email and search and social media like Facebook and LinkedIn some changes might be happening, right? Or maybe I'm investing lesser in some of them. Talking about investment, the fourth B would be promotion. So first thing we should check is whatever promotional budget I had last year do I have at least the same budget this year? Because if I'm having lesser budget then I may not be able to easily meet my targets, right? And maybe the return on investment from some of the media like for example, let's say Facebook has not performed as far with expectation. So in the advertisement world that's called ROAS or return on advertisement expense. So one thing I would like to call out is we are studying all of this keeping in mind that the attainment is 90%. The exact same exercise would be done if it is 120% but the decisions taken will be different, right? Because if it's 120%, you want to see what is the success maybe a certain product or a certain place is doing something very nice that you would like to share with other products and places as well, right? When, trust me, when you do all of these descriptive points, data points you collect into tables like this, my dear friends it becomes a very powerful exercise. Let's say that when we did this exercise we found out that offline channel, even though both channels we are investing as much as we did last year offline channel now has 15% lower returns, right? So the revenue that I'm getting is lower and that's how we decide that the return on investment off offline has reduced. Collectively the decision we take is let's optimize it. Nothing can be optimized at a channel level or a region level is one theme that we've been seeing throughout, right? So we decide let's go to the most granular level. So there are 25 cities in which we've been investing so far and what I'm going to show you next is an exercise we could do where the x-axis is your expense and the y-axis is your revenue lift or the increase in revenue that you would expect from each of these cities. This is what you really get. So on the x-axis the expense, revenue lift on the y-axis and we do a scatter plot for all of these 25 cities. If you watch carefully there will be an invisible linear line that you can pass through most of the topics, right? So let's do that. Let's do a simple linear regression, right? What you will be able to see that while most of the points do fall within the upper and lower bound, there are five points that don't. The ones at the bottom marked in red are really underperforming. Why so? Because the amount of spends that we are spending in them we are not getting the same ratio-wise revenue lift and quite the reverse is happening with the ones marked in the green oval. They're getting an excellent return on investment from them. So what we decide is without investing a single extra dollar and using the simple 2x2 matrix, using simple quadrant analysis, what we can do is reallocate some of the funds from the bottom right bucket to the top left bucket and that hopefully will make both Karli and Mahesh all the stakeholders happy. Now let me call out that it may not always be a simple pen paper exercise like this. Sometimes you will have business decisions why you cannot reallocate those kinds of funds or why you are okay with lower return on investment from certain sites and that's okay but everything has to be either backed by data or by a business reason, right? That's the whole purpose of it. So hopefully with this case study and when you implement all of these things you will be able to make your clients shine in their client size because that's the ultimate victory, isn't it? Now that's from a learning perspective. Next what I would like to do is define our next steps together. So the first thing I would suggest is if you haven't already go through the various business frameworks because they help the analytics process, ML process, anything that involves projects that involves data very well and the second thing is use this big useful surprising checklist to filter out different insights. If you are a student what I would like to suggest is practice all of this in whatever next projects you do. Whether that's on Kaggle or your academy practice them and discuss a lot of different problems with different data professionals, right? You need to have exposure to different domains, different types of scenarios so that you are more ready when you enter the corporate world. The second thing that I would like to suggest is for existing data analytics professionals, right? So you are a master of data as well as technology already, right? So try to spend more time with your stakeholders and clients, understand what they're going through at a daily level, what their goals are, what their priorities are, right? And what are their pain points? What is it really that's troubling? And that becomes your big business problem statement to start with. After that you have everything that you've learned today to actually apply frameworks and help them with solutions. If you're an existing business professional like a marketing manager or a finance manager, you already know what you're going through and what your goals are, right? So sharpen your decisions through these kind of frameworks and checklists. And if you're having a larger team, I would definitely recommend have a good mix of people who are strong in domain, in technology and in data and then help apply all of this. Finally, if you're an analytics leader, of course, share all of this with your team but also pick your change agents, your early adopters carefully. Give them as a pet project so that they can create the revolution that you want in terms of problem solving. That is all from my end in terms of the session. And thanks for being a wonderful audience. I'm happy to take question answers.