9. Making Data Science Work

Some of the best projects that I have worked on. Have not involved any modeling. Have not involved any sophisticated stuff, but they have been simple cuts of data, but with very high impact, as long as you show it, right? You don't need to build the rocket ship out in the first goal, but you can at least show that, hey, there is this part of the rocket ship that is built in. That could be useful for you in this way. Hello and 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 Church that Blogger newspaper, columnist book author, and a former data and strategy consultant at currently head analytics and business intelligence for delivery. One of India's largest logistics. Companies. You can follow me on Twitter at Karthik s that is Kar Phi.

K s and read my blog at no Intruder.com. That is n Over. N th you be a.com all opinions expressed in his podcast belong to me. My podcast this and I do not reflect the views of any organizations. We might be Associated. Nothing discussing this podcast should be taken as Financial or legal advice. We are now in 2021. Everyone wants to do data sites. When companies raise money. It's common for them to say that they have used the funds to invest more in data, science and analytics.

Some other well-funded companies are accused of holding. Doesn't this back in 2006 during a bull run and crude oil prices mathematician, had said that data is the new oil and that's now become a sort of self-fulfilling prophecy. You're just investing in capturing data and hiring data scientists doesn't work. A lot of effort goes in to make sure the data science teams. In actually be effective. Today's guest is unknown, Krishna. Anuj was an early Employee Of Music.

Mama one of India's largest analytics, Outsourcing companies data, he went on to be a co-founder of the math company over the course of his career, and it has left and delivered and it's of analytics and data science projects, and by analyzing data from those, he has figured out a thing or two on what works. And what doesn't in terms of making data signs work for a company. Let's start with the how business and data science sort

of interactive mean. This is a sort of slightly old view that I have. Which is that like the business guys, Kenneth treat the data that is in one of two ways. One is that like they think the data people are gods and they whatever they say is the gospel in, you need to listen to them and the at The Other Extreme you have like you the the business people think. Like these datasets are doing some random math, which is not relevant to the business. So why do we need to listen to them?

And if either cases, if you think about it, the business is not able to get the full value from the data. So start with like, is this is, this is my view, correct? Is it still happening? Is it Universal? Or is it sort of certain types of companies, where it happens in? Yeah, and what do you need to do to their sort of integrate data and business better?

Yeah, I mean, so I think that, you know, I think a lot of it has to do more with how we'd how data scientists are, you know, that the data product itself gets delivered to business, right? I think because of that, if you, if you think about how that that telling really happens and if you make it more math, more Tech, and you don't do playing speak business, speak businesses, tend to business. People tend to either twist to the side on which to that side.

In either case, as you said, it doesn't become useful in terms of organizations themselves who, you know, kind of go through that. I haven't seen this as much in organizations which have sort of grown with the data at its Center or text at its Center Sovereign more newer Edge organizations. I think like if you take the Amazons of the world, I don't think that they think that differently between data and business because for Them.

The data gives them, the is the core of their value proposition in some ways. Right, but if you take some of the older organizations, it becomes a big challenge because they're not used to thinking of data, in a particular way. They're not used to using data in a particular way. So, so there it becomes more of a challenge where where, where businesses sort of tend to tune out in a lot of ways. When, when, when, when the data is come to come to talk anything. Not.

If it is just articulation. I feel. I mean, I think the, the if the data product is presented in a way that is simple, that is usable that caters to, you know, a layer of translation because you do need to translate what has happened in your data science, sort of cooking to. What does it really mean to the business? If the translation layer is not very clean. I think in a lot of cases that That sort of you things to happen. So computer translation there.

It says they are they whose responsibility it is a tiny or to phrase it another way, right? I mean one way the some companies took it is like get a bunch of math, the agent didn't do the data and then they whose responsibility is it to sort of take the Data Insights? And because I have seen like client presentations, which have the this head, this the value, the set that we value. And everybody was like black.

I mean, so obviously the door, so somebody needs to take the responsibility who can understand both sides. It's okay, take the P value. And then, then the business,

like, okay, this is what the data is telling us. It's also how, how does it sort of layout? I mean, original distant, different organizations have dealt with that very differently, right? So there are some who have who have a specific person who, which is a role called the analytics translator.

If you will, who is going to do that for the business and from the business to the, to the, to the analytics folks, right in our The case, for example, I think that we have been, we also believe that unless the data folks themselves, learn how to do that, at least a little bit. It becomes a pointless exercise when you build a model and it doesn't get anywhere kind of

thing, right? Different folks have done it differently, but I feel that the the important thing is to understand that, you know, you could bring in folks for playing different roles whether it is the same person doing that or not is a different question, but there are different roles that need to be played. And that identification identification of those roles is extremely important. You need somebody who's going to think about the cust customer or

the business user. And I'll figure out what it is that they are really looking for. What is it is that they would like to hear and what they don't like to hear and make sure that someone is constantly playing that, I think the owner of that data science team and that that function, if you will and whoever that might be it could be the, you know, it could be the CTO. It could be the CIO. It could be, it could be the CD of whoever that is has to be very Cognizant of the fact that

that needs to kind of happen. And if it doesn't happen, then everything sort of goes for a toss. So identifying those roles and then based on those roles, figuring out who's playing them and making sure that those roads are actually being played, but the owners to be honest, is always going to lie on the people who are creating the product, right? Whatever? That product. So the owners can never lie on the business. I keep going back to this

example of think of your think. In consumer right at the end of the day, if the in consumer like a like if you don't use an app, you are not using an app. It can't be your fault. You're not using an app ever. So I think the same applies for business users also. So what do you say is that the data guys need to sort of, like, I mean, it's important for databases sort of know some business.

Learn the business context solve business problems rather than solid mats problems because I think that's one wrap that it's easy for data scientists to get into right - you just go off into the exam, different levels of geekery, depending on what what problem you're solving and like that can sometimes temperature. Most elegant solution need not. He produced a good business impact in some a gesture simple average somewhere might do the trick rather than building some

involved models way. Yeah, absolutely. I think that this is something that I've seen a lot over over over time. And, you know, I think a lot of times we like Tech Community, right? Whether it is a data scientist data engineer. Anyone we tend to get lost in our own. Complexity of our own sophisticated problem. Sorry, sophisticated solution. What kind of model? What kind of new technique are? I am I going to use things like that, without completely from beginning to end focusing on?

What is really the problem that we are looking to solve and work, at least from my experience, you know, just thinking about it with a sort of a design lens if you will. Also looking at, you know, understanding where the source of the problem really is understanding the real solution that somebody's really looking for. And then tailoring, what? You're sort of data science solution. Should be needs to happen which lot of times just doesn't happen.

And yes, you need to understand the business for it. You could get help and somebody else provides that business context, but somebody really He needs to always do that on a consistent basis. Otherwise, it just completely gets lost in, in the, in the, in the math and the tech of it. I think that the that also applies on the other. And if you think about it, as I have a lot of results, right?

And we also tend to enjoy the results that are coming out and try to, you know, marvel at at our own sort of magnificence in figuring. Not that result not thinking about. Okay, fine. Tied back to the problem that we have solved. How do you make sure that, you know, going back to that point and communication? How do you make it simple for them? How do you make it a easy for them to understand and read it? You know, how do you make it

visually appealing? How do you make it, you know, even, you know, something that they can quickly, touch use things like that, right? Thinking about it from that kind of lens is hard for, for of us to do. You could really, you know, one of the things that you could also do is, you know, for example, some of the things that we are doing is we are bringing in actual designers who are going to think about it, purely from the perspective of what is a customer who want to look at right?

And what would they want to look at? What would they want to kind of see and feel touch. And, you know, think you thinking or things? Like ux, for example, which I'm sure a data scientist, wouldn't think about, right? If you are putting out a Patient tortured, the presentation kind of flow mix. Make sure that makes sense. Things like that. I think it's a very critical part which is often gets lost in the in the in the complexity of the solution. So, how do you think of the team for this?

I mean, like that's should everyone sort of the have some mini meant of communication in the more, you try to bring in Specialists for the communications within be the bridge. You're like, how do you solve this problem? I feel that there is a the the

there are only very few unicorns will be able to do everything, right? So I think that's that's one but there are there is there needs to be an appreciation for of everybody for everything whether that appreciation did not be deep but there needs to be a basic level of appreciation for the other aspects of you know, what what needs to happen.

Even on the engineering side. For example, a lot of times I find that data scientists don't have Appreciation for what it takes to get the data to where it needs to get to, but you need to have that kind of appreciation. Otherwise, you'll kind of, you won't understand the challenges in the complexities that they have to deal with in your solution. Hence won't be designed to deal

with those complexities, right? And on this side in terms of design and whatever not it is the route that we feel and what at least we have taken is, is more on the build basic appreciation. Get smes. That's, that's the route that we have taken. So we have people who are doing particular things, as you know, that's their core, but they have an appreciation for the other stuff. How do you do it in a mean?

Your sister sort of thought they'd see your largest laddish organization serving several clients into on so you can connect with their grief so that I think it will fit but to the with a small team, I guess it had a rate. I mean you'll need to have some unique or somewhere else. I don't know like I'm struggling with the dog. Yeah, I have absolutely I think that there need to be some

unicorns. I would also say that the needs to be Trini that Is really that that needs to be enabled for some of these people I've seen in, at least in my past, that, you know, doing something like a design, thinking training. So that people understand how to really think about it, is extremely critical because a lot of times we just don't get exposed to things like that. I think that the traitor the exposure is the greater, the training is even for small

teams. I think that's extremely critical because it's just at least it up levels. The work product. Wore a into a into the into just that next category. Now this coming back to this communication regular. Having register start to the board broader slate in terms of like how do you mean? Like, let's assume you have it. Some cool model, whatever. The model is. It could be a regression. It could be some, some neural networks are all, you know, but

like how do you sort of? Like, I think one of the challenges that data science cases, I mean again, let me know from getting this wrong, but on my personal experience is that trick is communicating your results. That's equal to, you can communicate whatever your the inside your phone. Like. How do you like maybe the data itself as a story? How do you sort of communicated to the to the business? Like it should you abstract steps to do? So would stop to them the way that they carry?

Are you go words from? It goes back to the translation, but I think that I heard I read this somewhere that basically humans, basically can only understand stories.

Efforts. And so it's I guess important for you to tell a story for you to contextualizing in in their kind of perspective to I think they need to edit is also very high if you're not able to edit what you want to present properly, and make sure that you cut out stuff that really they don't care about it becomes a big challenge, right thing using

examples of you. So say you've built a personalization model of some sort identifying, a cohort of customers and showing the kind of personalization you really talking about as an example, bringing Comfort to them to say that. Hey, you know what? I have the model and all is doing what it's doing, but this is the result of it. And as you can see this meshes, well, with your business intelligence that you have built up or you are thinking about your intuitive sense.

So you can trust that. There is data. So that is a flow that is kind of Happening Here. Bringing out some of those examples bringing out metaphors. Bringing out stories. I think is a better way of communicating the results rather than thinking of it as I will talk through the process of how I built my model. The wait, okay, so it's basically need to be covered in

some ways just to replace. What is it about going backwards said, it's like taco dissolution and then like, only they had like really interested in. So when you picked up over the brush because it crosses doesn't necessarily add value to the business. Yeah, I think, what they caught what business are support more, is that is your process robust. Have you dealt with all the, all the possible, you know, said data, cleaning challenges that you might have had.

Can I trust The data that you have used. These are the questions they would have and you don't need to go into depth to show them unless they ask for it, as you said, but it's not a necessity for you to get into depth to show it to them right there because I did sometimes.

I mean, I guess especially something like data science like over selects for people who are like sort of interested in maths interested in processing interested in doing cool things like like they Instinctively forcibly over index on, on that rather than rather, than on the, on the story story, eats in Vegas. Yeah. Absolutely. I think that is why I think he designed us need help to be honest and bringing in that hit from as much as many corners as possible becomes critical.

So in a small team, making sure that one person is constantly say thinking about it or making Show that at a bare minimum level, you put some standards process see saying that, this is how you need to be at the minimum showing your presentation. These are the things you should not be showing. I think some of that helps, you might say that like data scientists need help.

So who are the kind of people who can help the data center is, are these like buying by the 10, but everybody wants to be a data scientist are like, sometimes you might have a higher data visualization engineer. They'll want to be a data scientist, you you, Hire somebody to hide over here for whatever reason like this also the some sort of like something simpler than others and so on. So so how do you get help to data scientists and make sure that you can get people who can

actually help the data sent. I mean, I don't know for a smallish team, but at least like I can tell you from my ideas. My own inexperience is basically get designers clean. They come with a so it's like you talking about orthogonal perspectives, right? It is completely orthogonal perspective to a data science perspective and making those two mesh is probably the help that data scientists can have need this probably but that's only one example, and that's very

specific to our case. Maybe, but if you think about it, I feel that, you know. We were talking about it earlier but bringing in more appreciation for it might help bringing in a proper training program for, it might help bringing in some some base level practices can help. But other than an external entity doing it in an orthogonal manner, it's only the up, leveling of the data scientists themselves, which can help what. And speaking of data scientists

Zips is set. I mean, like I have a lease space on my anecdotal information, like some kind of a, some kind of a curse distribution to for the lack of a better phrase site in terms of like everybody wants to be a data scientist. I didn't make data save because of that data scientists think that they are the coolest 10 like they are like as I read in some log musically.

I mean that's a Blog have quoted in some five different places already is actually data scientists believe that business intelligence is for Lexington's increase and and then like, so I did this is impression that the people who are doing it's a business intelligence or analytics. It's not as cool as data science. The data scientist want to work on the latest cutting-edge technology. And so, how do you, how do you sort of like when somebody comes to that kind of a mindset?

Like, how do you up level them in terms of better communication, better, appreciation of other domains and things like that. I think you like really like I think we've been dealing with this question or on business

intelligence versus cool data science for so long now, right? I mean, I'm sure you've seen it for so long also, it's it's a fundamental problem of not understanding the objective, I feel and you know the focus at the end of the day. It's like it's a whatever you're doing is a tool to solve a business problem. That's what it really is. Some of the The best projects that I have work.

Done. Have not involved any modeling, have not involved any sophisticated stuff, but they have been simple cuts of data, but with very high impact, as long as you show it, right. And tell that story right? Which made made customers like millions of dollars or whatever it is. Right, but that perspective that look at all all of what you're doing. Whatever. You're Doing, let's back to the

business problem itself. And it does like a tool with which you are using, doesn't matter as much as tether. You actually solved it or not. I guess. I think that's that's the that's that perspective. Sometimes goes missing and you need to drive that perspective as much as possible into into people. So that then they realize the value. But some, some are to be honest. Extremely just fascinated by the by the complicated stuff and you can't stop.

Up that. But as much as you can drive the perspective of here, you know, what? If you solve the business problem, you will go to hire like lengths and you know, you'll become so much a much better sort of, you know, you'll have much better growth as long as you can drive that perspective. I think people in a see it and understand it and and as they experience it themselves then they like it. Everybody likes things when when take many gets used. That's just the truth of all of us, right.

So yeah. So when they see that okay usage as happened and because of usage happening, something has happened. You have a sense of sort of pride in your work and that sort of leads to leads to things, but going back, going back, pushing down. The message of the business problem itself is, is critical. I don't, I, I have always had an issue with this with this complicated stuff. Versus business intelligence stuff and it's just never gone away. Yeah. No, I guess.

Hi, Abby, I completed a human in my experience. I mean, I've only been doing this for 10 12 years now. Like I've never really I don't robots that are single case where I had a complicated model that delivered like massive impact. All the, the best part is always really taken over eating somewhere again. Something I quote often, like the best machine learning method is division. Lik take 90% of your work gets done with average estate. I don't like which is also

linked to having back. When I was, ER, there. I need this country take people would ask me. What kind of methods do you use in your analysis? And I would be the in 90% of the cases. I use averages in the 90% of the remaining. I am use regression and maybe beyond that. Like, it's the thumb. You almost never get there a quite But people have always focused on that last one for today. What is the cool stuff that you're doing? What are the cool stuff?

I am doing? And why did I sign up for this job? Where you're just asking me to find cuts of data rather than sort of doing some models and stuff. So yeah, I mean, yeah, I think there is a there is a, you know, that is a I know mindset issue in some ways, I guess and but again, this goes back to what is the business? How are you gonna kind of solve it? I completely relate to what you're saying. The, I've always had an issue with the same complicated modeling where people ask me

over. What are you going to build an RNN and you know of that and it's like why do you really need an RN M to do this trick? Maybe not. Right? Think about that, if you can get the answer quickly and you can get the answer right then why would you need to make it make it any more complicated? Get those revised. We have sworn somebody recently told me they were like, okay, you might be able to solve this problem like this, and this might be more impactful for the

business. But for my CV, I think that is the, it's important that I do some more machine learning and things like that. Yeah, that's true. That's true. Yeah. At the same time. I don't want to say that there is not a time and a place for doing those things. Obviously there is but it cannot be the it cannot be the base of everything. Needs to be contextual to the problem, the the the data that you have so many different things, right? What will customer even

understand also matters. If you build a very complicated solution and you can't explain it well enough then really doesn't make sense. Coming to the customers rate. So so we'll do, do you have cases of customers asking for like complicated solution or do they just want the problem solder? And you like? It depends on really obviously you have a spectrum of of different folks who ask for different things.

But lot of times, I think business users only care about a, I need to, I need to make this decision and tell me give me information so that I can make that decision. Ideally, give it to me very fast and also make it easy for me to understand. I think, if I look at it like 60 to 70% of, of At least the people I have worked with our would Fallin Fallin that kind of category.

There is going to be some people who are either dealing with extremely complex problems or are trying to push the boundary of, you know, what is the next thing that they can do or are just enamored by a complicated, you know solution? We will ask for that. But I would say majority still people who are looking for help in making their decisions and I want to make sure that it gets done fast and wants to make sure that they understand what is going on. I'm generalizing the salad, right?

So the next thing, you're a business person who is engaging a data team. I mean, either within your company or as a, as an outsourced vendor or whatever, like you're going to business purpose, in engaging, the data, viewer, your job is to kind of take the insights out from them or maybe get a product from them or something like that. Very question for you is like, what do you, how do you, how do you paint a wall more? Do you as a business person need to do? Take the best value out of the

out of the data. No, so that's a tough question.

I think providing a lot of context matters as much context as can be provided matters. Demanding understanding of problem, is critical. Tell me what you understood of what I want and asking for that is, is very critical. Maybe even asking for water sample solution, would look, like is very critical, get that. Get all of that out of the way in the beginning. So that, you know, further down the line. You don't have surprises as a business user. Right?

I think it becomes imperative for a business user to know that there can be a tendency to get lost in the solutioning and to to keep that bounded and keep that tight is probably very, very critical in for a business user. The the beginning in, The beginning, it becomes extremely important. I've seen a lot of cases, where things fail because not enough information was given by the

business user or not. Not, I mean, nobody took the effort to make sure that an end output was presented as a mock in the beginning. So expectations were very different in terms of what I'm what the business is going to get worse, is what the data science team is delivering ending. Small things like that, help a lot. Right? And does the business person need to know any little bit of data science at our statistics or whatever.

They just to make sure that the data team is not bullshitting them more than that, more than knowing statistics or data

science. I think, you know, applying their their own business knowledge and asking for for examples, helps, you know, saying, okay fine, so, can you Tell me, I mean, for example, when we work with say B2B customers who are selling to knowledge organizations in the u.s., You know, they know their accounts say in and out right say they know that this kind of a large large say conglomerate buys these things from us and you know, to ask the question of

how much our, how much is your model saying that they are going to buy the next month and Obviously have a sense of whether that is even ballpark right? Not ballpark re-think checks like that become more important rather than knowing any progression or data science or anything like that. But wait now, I mean like learning things a little bit. I mean we were talking about make you spoke about a couple of cases.

In terms of flick business. Failures data are not specifying the problem properly or like they're being and expectations mismatch. This about what all the other cases where they sort of data science, teams of projects field

at why L die, I'm sorry, like through your journey. You would have you would have had your share of some projects is would have worked out. One day. What happens? What happens in the cases when you signed it with the way? Why do product project scheme? Yeah, so in some cases it is alignment where, you know that there is there isn't enough alignment between either the stakeholders in the business themselves.

Right? What someone else is expecting versus, what some other person is expecting, is not the same thing. In some cases. It is alignment between the teams, you know, say there's a

data science team and there is a business team. and if there is inherent mistrust from the business team through the data science team can kind of fail, sometimes it's, you know, people not, you know, not providing like the data's data, guys, not providing enough Comfort to the business teams saying that, you know, this is something that you can really use And this is something trustworthy. I think that, that, that translation becomes, if that doesn't happen properly.

It kind of fails, not thinking about end output and usage of in doubt. Put enough, you can, I think we talked about it before. You can create a sophisticated model. But if, if it doesn't deliver the in doubt, put in a usable format. It doesn't make sense. Sometimes not being scalable in your solution, if it doesn't integrate with You know, Downstream processes, then anything that you build is pointless, so that that could be a reason.

Sometimes it is not making things, simple enough, and not making it an enjoyable process. Right? I think that's critical when you are working with people. You have to make it enjoyable that for them also. And if that that is a very, you know, friction filled process, then it it will fail. L. I'm reminded of one of our assignments about some 78 years back. We just reattach it. So, what I did was I actually, like, I remember sitting down with the ab is one VPN 1gm, I think.

And like, actually, taking them through the data, taking them through the model saying the, this is what the clustering put out. And, let's take a few sample customers, and this is what they look like and things like that again. And at the same time, I think they were working with another larger established and McNair and they were like, okay the they quickly figured out that they could use my model because I had given them the comfort of

like the actual them who are. I mean, we did speak about it a little bit career would not going through the process and stuff but I did show them some some facts of the data. I showed them some sample, customers its own, and I think that just generated this tens of comfort for them. Yeah, I think you know about one thing. I'll also add to that is I think just the going off of the examples are closed set when you're working on it.

It's also important to keep going and showing something, you know, there's also this tendency that in these kind of team sometimes, what happens is that you tell me what I have to do. I'll go build what I have to build and then I'll come and pass it back later, but that doesn't work. They, you need to involve them in that process and Keep going and giving them pieces and show how the, how the sort of the house is getting built and not just show them like, hey, here's

the house. Right? So I think it's, it's that process has to be very, you know, Comfort field and enjoyable. Like like you said, yeah. So I think that it comes over again to the Exotic proper alignment and ownership and stuff. They didn't. Yeah, alignment ownership and and also trying to yeah. Yeah, I mean, see if there is a base level of alignment issues. Then you really, you know, that it is set up for failure from

the beginning. But but if, if there is a little bit of an open mind to say that, hey, you know what? I'm skeptical, but you show me stuff. And I might, I might get convinced then you have scope for doing things which, which gives you can take to the next level. By having their white but I think you mentioned about how like we need to serve the pad mock-ups in the beginning you need to show that I could the insinuation and then sort of go backwards with it.

So I was reminded of a in some people in writing we call ideas are formal term, would be quality reverse, pollute format. When you start with the end, you start with the headline on top with the main conclusion. And then die going to how you got, too intense one. Don't we don't keep the suspense to the Indeed absorb this again, really doing the data project is again sort of Click one thing where the worst Bollywood for my

pulse rate you start with. This is what I want to give you and then make you fight these like some vaporware in there for now and then you will feel it in the in Lake will the modern and also CCS. Yeah. And take them through those right relations, you know, look at the end of the day the the the this field and this is the kind of projects in the work that we do. It is always Who is going to be titrated? You can't avoid it. The iterations are, are, you know, they're just going to exist.

So if I durations are going to exist, it is better for you to kind of prepare them for it. Make sure that they understand what you're aiming for. Make sure that they understand, you know, what what what challenges you might face and prepare them for the right reasons and take them through the Journey rather than just, you know, dumping one thing in the integrative. Yeah, I think so.

You have seen data Journey since I think the term data science got coined in 2008, which is a around the time. I think you enter this into the sea and so on. So so how was the investing world over the last decade or so and like, oh wow, and what, how has what excites people like, could be their data centers, could be the business days,

could be clients. Like I'll have the expectations Change Daily. Wow, word and then in the middle you had this whole big data think it's one how well things changed in the last. Dozen of years. So when I started the actually, the term data scientist didn't exist. I remember that. My first designation was business analyst, so that was quite some time back, but I think a lot of lot has changed. People are a lot more, like, let's talk about it in groups of

people, right? So to if you talk about, if you talk about folks, for joining this field now then they have a certain expectation of what they need to be doing. And we talk about Of that what they think school. And what they think is things that need to need to work on and things like that. I think there is a lot more knowledge for them. Also. They know it's lucrative right? Fundamentally for them to be a

part of that this industry. I think that is that that knowledge is very well ingrained. I think in a lot of people trying to get into this field. So if I mean, ideally I would like more of them to be excited about the business problem itself rather than And rather than the, the, you know, better than anything else, but the techniques and the tools sort of take over a lot of times that's on the, that's on the, you know, folks joining the field

perspective. But if you look at businesses, businesses also have gotten a lot more knowledgeable before you could go and say stuff and you know, they would they would listen, they would listen to you because they didn't know much better. Then but I think businesses business users. They have heard they have read, they know a lot more. They have tried. Maybe have failed, maybe has worked. It's just, it's a, it's just that maturity curve that they

have also kind of gone through. And I think the expectations now is that he, you know, what, keep it simple for me. Do it fast for me. I don't have like, you need to, you need to come to the table. It's something that you might have already kind of done before you have, you know, bring your knowledge to the table in some form or fashion. Now, you know, so that you don't start from scratch, do it fast. Do it, do it. Do it simple.

And make sure that whatever you do you leave me with something which is scalable and usable. It can't be that you know, you you solve this II, get you to solve this problem once and if I get this problem again, When I can't use your solution again, it has to, it has to scale. Either to other places. I can use it to or two other points in time. I need to do it in, so it needs to be scalable.

It needs to be usable and I need you to do it really fast and but make sure and keep you keep it simple. So I think the expectations have gotten so much more higher and that's a that's a journey that I guess every industry kind of goes through and and this industry is going through that same. With expectations getting higher is the output also like, eating up, or is there a sorbonne this milestone in terms of what people expect? Anyway, what's delivered? I don't know II.

Think it it will keep up. I think the fundamental economics of demand and Supply will play there and somehow it will keep upright. But yeah, I mean, it's not like it's, we could test and it's everywhere and everybody's doing it. But, you know, there is enough and more effort in, in a lot of places to, to make sure that it we try to keep up with that expectation and demand.

So, you mentioned one thing about how, like, nowadays the businesses want the answers quickly and you need to take something from your brother from your previous experience and things like that, and so on, right? So I think this is one of these classic care if dichotomies, that businesses need some output quickly, but typically the especially if you are trying to get into the modeling research each, I never think the solution always takes time to be.

So, how do you how do you got a kind of bridge the gap? Apart from? See coming up with some decent margins on using this. Some Leverage. Is it be through there? Yeah, I think one is obviously it's important for you to however you can you uh need to bring that to the table, right? That experience in terms of say preset models, reusable codes, whatever it is, right, how can

you bring efficiency into? It has to come, but I think the other part of it is, is is, is, is basically, I trations, right? So you don't need to be. Build a rocket ship out in the first go. But you can at least show that, hey, there is this part of the rocket ship that is built in. That could be useful for you in this way and giving those bits and pieces over time and eventually showing the rocket

ship becomes kind of critical. I feel because otherwise, you know, otherwise, it is just update that that just happens creating MVPs. I think is critical and goes back to design thinking in I miss that. You know, how do you how do you do prototypes quickly and show it and and show how it could be useful? And then that kind of shows the direction of where you need to connect it to one thing, which I think I wanted to ask you have

which we haven't covered. Like I think we've spoken a lot about the relationship between business and data sets like the give, but the other thing which sphere it's probably gets a literally less footage, I think. Is the relationship between taking data that data sites or analytics? So how does use a horror? How do you see the attraction between taken data science? How virus has evolved over time and like what are they sort of put the impact practices on?

Yeah, I think that right now at least taken data signs are extremely integrated. I think that it's very important for them to be integrated as well. I think going back to the solution building that I was talking about, especially. You have large amounts of data, then Tech becomes integral anyway, but even in other cases, say you're, you're bringing in data from so many different data sources for to manage that and to set it up in a way where it

can be used in data science. Also becomes kind of critical so Tech and data science are actually getting closer and closer every every single day. Right? And it becomes very important for for, for For dear scientist to have an appreciation for what, what, what the tech side brings to the table? And for text, I to understand what data scientists need. And I think if you look at most organizations today, the there is a there is a close relationship between both of them.

And I only see that relationship becoming stronger to be honest with you. I don't know about whether you know, whether this will sit with that or that will sit with this and all of that. But, but at the end of the day, They need to work very closely. Like we had to build out data engineering team and a tech team ground up to you know, to make that happen. Otherwise, we knew that we would kind of fail in whatever we're trying to do.

The other part of it is on the other side of tech which is more like product depth, you know, that also becomes very important because if, if you are not able to do, you know, people are expecting products, right? This is the world we live in. In is hyper personalized but also like apps and all of this and people expect those things from from from, from whatever

your building. So if you're not, if you don't have a clean productive kind of setup, then that also kind of becomes a challenge because then you are not going back to our earlier points. You're not presenting, whatever you're doing in a manner that they will want to use. So, so yeah, writing the tech tech and did data science. I feel are getting closer and closer. And more and more integrated. And we need to get more and more

integrated. If you are if you're going to deliver to those expectations about ability to that adding technology over the last 10-15 years is completely Embrace. This role of the product manager to sort of like a physical big. You go between between business. Take the design, Etc. So do you use either dies or other similar grown in data? And if so, like what, what, what does that kind of role look

like? Yeah. I think, you know, there have been attempts to kind of Define that Troll. But in a very narrow manner, I feel like I remember that. There was a whole article around the analytics translator and why that person is important to you. Things like that who's kind of it's basically the linchpin, who's tying everything together, right? I think that's kind of what, what what, what you're referring to and I think that that role is, is there.

Usually what ends up happening. Is that the person whose front facing with the business, please? That role but it could technically be anyone playing that role. It could be the data science person playing that role. It could be the tech person. Playing that Ron, it would be a design person. Playing a tank. Could be anyone usually where it lands is is the is the person front facing to the business

ends up playing that. I'm sure you are playing a role in the world in the world that you are in with the business. And that that's probably the easiest because of the fact that at the end of the day, as long as the business is happy, then you're good. Good. 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 vodka. Has Spotify or wherever else?

You go to get your podcasts? Once again, this is Karthik signing off. Thank you.