From Raw Data to Real Business Impact: Mastering Power BI and Microsoft Fabric with Thummalacherla Krishnakanth

Yeah, welcome everybody. Back to the MC65 podcast where we talk about all products, stuff around the Microsoft, the future. Yeah, and interviewing experts today's guest, Chris Laka is a Microsoft MVP, Senior Power BI developer at Questblow. Through his career, he has divided, and the price analytics solutions for organization, including Exxon mobile, Rolleroy, Siemens, Gamza, I think business does for broad data into meaningful insights. Real exploring everything from Power BI architecture, DAX and optimizing fabric, snowflake, data bricks, AI analytics, career advice, and so on. So yeah, welcome, Chris Laka. Thank you for being taking time to be here on the MC65 podcast. Yeah, thank you. Thank you, Malcolm. Thank you for this opportunity. Hi, everyone. Thank you for joining here here. So myself, Krishna come. I'm joining in from Bangalore, India. So overall, I do have a seven years of experience and currently working for one of the service based firm, organization for one of the oil and gas projects. And I'm also a fabric super user. For which I got the recognition for the data community, for all the contributions. So yeah, that's about my short introduction. Yeah, for listeners, you don't know you. How have you started your career and what especially brings you to the data world in the Microsoft ecosystem? Yeah, so initially I started my career as an electric engineer. So that's where I used to start my career. And then I already from my childhood, I have the weaving of the rocket launches basically where I started my childhood from the Indian Space Research Foundation, a boundary where all the family of ice row, where the rocket launches will be keep launching on. So where my father also the part of the launching activities. So that's where I used to think with respect to analytics part, where that's where my during 2020, 2020, 2020, I started thinking towards the analytics part, how can I come up with a gap? So can I come up with the visuals? Like how many launches has been getting in order in each year, year, on year? So how many are getting success? How many are getting failures? So that's where my thinking has started towards the visualization from the electrical engineering. So that's where I made the transition. Awesome. It's interesting. Yeah, a lot of people say now with a eye, power, or other, a lot of people say, that are people say, excellent depth because we have power, be eye now, a lot of people say, okay, power be eye that because we have a eye. What's your view on this? Yeah, yeah, because so most probably I don't think like, we need to be always with the basic parts and the fundamentals with respect to the power be this time question. I agree partially with things, what you're saying. So so coming into the AI picture, it completely takes a lot of time down the line if we take it. So it takes more time where we can see the incorporating of the AI things directly in the port. So as of now, a developer can successfully learn if he or she is interested towards working with the public dashboard development. Yeah, I think, yeah, power be eye it's for me, it's the coolest product I work, especially in building reports. It's cooler than I don't know. Yeah, SAP, SAP, Einstein and so on. But there's a new product. Have you have you also try out the fabric apps? Fabric apps, I didn't get a chance, Mirko, but I started exploring into the fabric workloads with respect to the data, like, Oh, where else real time. Oh, that's that's cool. Data flow, yeah. What will you say, what does power be eye better than traditional reporting tools? Like, I don't know, Excel companies use. Yeah. So mostly when companies, it comes like there are companies that are using still with respect to the Excel and the tab. So when it comes to the Excel, we can where we can able to handle not use your limit of data. So there will be a certain restrictions. So they can access to bring in only limited records into the Excel. So that's where we can go with the then when there is a huge volume of data, we can go with the power be eye. And it comes to the tab. There is a licensing part which is a bit more cost. So that's where when it comes to power be eye, for example, if I take a pro license, it will be around $10. If I offer premium for user, it will be around $20. So that's where the licensing part when it comes to the power be ecosystem. And also we are seeing monthly monthly updates, one by one. So that's where I feel the coolest thing to go for you. It is utilizing the power be application. Oh, yeah, updates. Microsoft does do a lot of updates and renaming and so on. What was the coolest update you see in power be eye? Yeah, in the reason in the recent point of time, I can see with respect to the date slicer with the broad with respect to the last grand up to last seven weeks last 30 days. So this kind of update will definitely will give the many end users where they can know need to every time ask for the developer. So by default, when we have those options in the top down, we can able to where basically the users can able to check it down like what the inside they are looking for. And you, I think you heavily use power be eye. So what is the, or what is the most mistakes people doing when when they're using power be eye? Yeah, so most mistakes which I noticed most mostly is basically when pulling the data from the data sources. So basically most of people there are directly pulling it into the reporting, instead of where they need to check for the data quality, how the data qualities in each column, how the quality is like for example, if there is a date column, there will be an actual data. And if there is a textual column, they will be having a certain date call date related values. So they, these are all they have to ensure that they are ensuring that data quality is accurate and they have to check for if there are any errors, they need to ensure that there are targeting result before pulling it into directly into the reporting. So they need to result all these issues in the parkory level itself, where we have the parkory engine inside the power by desktop. And also they have to ensure the accurate date types that they are giving. So by default, probably they will pick some of the other data types. So those things also they have to take care and then they have to ensure that right data modeling there before going on to the visualization section, they need to ensure the right data modeling is given between the dimension table and fact tables and also the cross filter direction also the matters. So they have to ensure always the direction is flowing from the dimension table to fact table. That means you do the downstream, but not into the upstream. So these are the certain tips where most most people makes the mistakes without following these and. So I would suggest these things to be followed so that they can ensure the values are which they are showcasing to the users are accurate without having any ambiguity. I think when we look at Lincoln and the power power BI groups, there it's a new, I think not really new, but I think a bus word is scalable power BI architecture. So what does it mean a scalable power BI architecture? What kind of architecture level you are just want to confirm it go. I read a lot of times as to people write this. What what means especially the scalable is is not all a power BI architecture scalable or is something special I have to do that I can say it's scalable. It depends actually depends on the project and the data that where team is getting so that's where the architecture comes into the picture and he where Arctic has to ensure how and what kind of licenses respect fabric if you have so what kind of fabric license we today either 64 SKs, stockkeeping unit or more than that we need to go depending on the work towards we are using. So that's where it depends the scalability depends on. It depends on each project. Yeah, I think there are I think we have the VM and looking the football games and so on and and yeah, I think also when we talk about power BI and architecture there are two teams two big teams to four favorites. The one side is the staff schema on the other side play the slow flake schema. When do you choose each. I lost your voice in between medical can you. Yeah, sorry, sorry, sorry. We have star schema versus snowflake schema. When do you choose each. So for example, when we have a denormal stables like for example, if I go with if I want to go with a star schema so where my dimension tables are all directly connected with the fact tables so this schema is nothing but where we'll be having a denormal tables. So which is also a best performance wise if we go, which is a best to utilize as a best practice as well. So when I want to when I have more number of tables like for example, if I have a product category, if I have product sub category in that case my product sub category table is not directly connected to the existing sales fact table. So product sub category table will be connected to the product category and then product category table will be getting connected to the sales. This is where this schema basically we used to call it as a snowflake schema, which is nothing but where we'll be having a normalized tables. So if your requirement is having more number of tables than I would suggest to have a denormal tables to make it like five tables if we have to make it as a three if you have three tables then I would suggest to make it as a one. So in the means we will be having a less number of dimension tables getting directly connected to the fact tables so I would suggest to go with the star schema as a best practice. And then we also have another yeah versus topic and there it's on the one side it's power query and on the other side is tax. What when you use what and why and what's the different between power query and tax. Yeah, so if we when we go with the power query engine level like for example if I don't want to unnecessarily create a colored columns in the tax level itself so where I can say that performance can be reduced. So if I can go with utilizing the power query itself then that would be very helpful where we can improve the performance rather than going to power query I would suggest to go with data source itself because whatever the queries you want to perform or you want to create few more columns based on the conditional columns conditional certain conditions based on existing columns I would suggest to go and write it in with the data source itself so that we can ensure that a report will not slow down its performance. So when there is a chance of where you want to perform kind of unpewitting columns when your data is not in a structured format so so those cases in those scenarios we can go with the using power query engine and where you can perform and if you have a certain tables of similar structure where you have five to six tables then you can go for upending the queries. If you want to merge two tables based on a common column for example if I have leak any column I want to merge table here and table B so based on leak any column I can perform the merge transformation. So if I have a similar kind of 10 to 12 date columns where I can see in the column level where I have the values so I can go with unpewitting the columns where it would use me the attributes and values. So this is how we can utilize the per query engine level in order to perform certain transformations according to our requirements from the data that we are getting in. When it comes to the DAX when you want to showcase certain metrics certain key metrics so that's where we can utilize the data analysis expressions which we have by default and we can showcase certain insights based on certain trends, Iran, here month on month using card visual KPI visuals. So that's where we can utilize the DAX analysis expressions. Yeah, that's good. What I often see in or one of my personal problems when the project become bigger power be I getting getting getting slow. What's your, what's your favorite optimization techniques here to to hold to make it. Yeah, so basically we need to we can open the performance and laser and where we can able to check from each visual level which how much time it is taking more. So that's how we can analyze by checking the how much time it is taking with each visual whichever visual is taking more time. Then we need to go and check in the data source. So what kind of how the column is and what kind of data we have so then we can accordingly we can ensure that switching of import whether we can switch from import to data query so that's where we need to analyze and we need to take certain actions so that we can ensure that the best performance has been achieved where end users will not feel that visuals are loading for more time. And a lot of companies I think they're doing giving all people all data is their one point. Yeah, we can say that the data set becomes too big. So that can become huge if they are adding on a day on a daily basis, Mirko, it gets huge. Yes, so they can best best ways to go with the live connectivity. So so that all the dead connectivity so it would be good if we hadn't have the data in the Azure analysis service so that we can utilize for the reporting point of you, if you say. And what would you think we have I think we have one side we have the use a management on power be I have to use a management inside the fabric when that you choose what tool for for as a governance to to use here. Yeah, governance point of view so now it is like most of the companies and auditions are switching towards the fabric environment in order to build the reports also into the fabric environment itself so I would like I would see like. Mostly they did know now we can see the data flow gentle as well as come so where we have a certain. When it comes to the licensing part I don't need to take a new licenses actual so if I can go 64 SKU capacity then I would start working on clearing of the data by getting my data into Lakehouse and performing data transformations using data flow gentle to. And then I can bring it into my warehouse and I can start performing the reporting I can start building on so that's where I can see when it comes to the power be I am fabric I would see. Down the lane we can see most users most organizations will ask us for to start working on with the fabric environment itself. And yeah when we talk about governance a lot of companies are especially tools like that thing we have in tune we have. Enter and we have view view what roles to display tools play when it comes to to data science on on the Microsoft part. Yeah, part of you differently they definitely give us the best security level when it comes to security I would see there is a good scope in the future in where we can able to have the good security when it comes to the data science where they can. And then you can also see the data that they are. And then you can see the data that they are. And yeah, I think, uh, actually, yeah, uh, Microsoft brings a lot of AI staff, especially co-pilot to all their products. Yes, how is AI changing business intelligence today? Yes, Miko. So what is happening is, for example, now, what are the report, I published it, right? So once I publish into workspace, now the end users as the capability to make the change. And they can able to book market. So whatever the published, which I, which I published the report into workspace, that it will not change. But whatever the prompts they are giving and for example, if we have, for example, one bar chart. So that if they give a prompt to change it to stagger call and chart, it will change. And it will give the insight what they're looking and they, where they can also be a bookmark, they complete report. So that's how I can see, uh, it was more into now, uh, where end users can able to change and whatever they look and feel they want to change and whatever the visuals they want to change. So they can able to change and they can able to save. So that's where they're getting more scope where people end users can able to look in for more business insights. Yeah, I think for two years, Microsoft, uh, I have seen, I don't know, a demo and they, they show. How they do, or how they automate ETL pipelines with AI. I tried the same way, but for my company, it's not working. It is, it is, I really, really helping data engineers or is it more a gimmick? Yeah, so it would be, uh, I don't, basically we need to be fundamentals good with the fundamentals. So as we see the genie in the data bricks, so it's a good, good thing that we, where we can able to analyze. So I would suggest to utilize the genie for analysis purpose rather than, uh, for the coding. So as the coding, if we have, as in when we have the fundamentals, so definitely we cannot be replaceable. So that's where I would suggest for the developers, like where we should always be good with the fundamentals and strong with the fundamentals, whatever the tool, whatever the, uh, platform it is. Definitely, we will not be replaced. Good. Um, uh, yeah. Have you or what did you think can AI now ride the most of the backscode or how good is AI in ducks actually? Yeah, sometimes if the problems which we are giving is not correct, yes, definitely. But basically we should be aware of what the context and what the requirement we are looking for. So definitely it will make a mistakes so that we should be aware of fundamentally what exactly we are looking for. So certain level we can go for utilizing the help from the, uh, the A tools that we have, but definitely it's always our responsibility that we should ensure that we should be having a good perspective towards understanding of the certain functionalities. Yeah, yeah. What did you think in the age of AI, what should data professionals do to stay relevant actual? So I would suggest not different completely on the A tools for your learning when you are learning or even your practicing, but utilize it for your analysis. For if you are not able to understand the data of columns, then I would suggest to understand ask the prompt to understand. So where what the data is happening, what the data is, how should I understand? So if you can ask these type of questions, then you will be coming to conclusion that what the data is, how the data is, what as a reporting developer, what I can bring the insights to the table. So as a best practice, I would suggest instead of going to utilize for the rest of the calculations, but I would suggest to understand the data point of view so that you will be the more better. Come up with a story detailing from the report. Yeah, and yeah, you have of this experience, this awesome experience. And you work for Zeman Skensa, I work for Zeman Sashmi, as long as you. And you also work for Royce Royce, I hope they give you the company car. It's one of the kind, it's one of us lines, so yeah, it's a good, good, good to work with projects. Yeah, it's a great experience. I think then we get the company car, you have to say, excellent. But yeah, you work for all projects, these big great companies. What have you learned on this project, what is the difference to, yeah, this enterprise companies to small projects. Yeah, so each domain is different. So Zeman Skamesa is working into manufacturing and roles for us into. And we have a craft, a craft domain engines and jazz at one into oil and gas. We into upstream downstream where we have. So it's a good, good. What I can say is very good experience, which I got from the data perspective and the business managers, what they are looking from their perspective. And it is great to learn the many lessons from the, especially I feel from the data point of view, because that is the biggest asset we have for every reporting developer. So I would suggest if you get a chance and opportunity in any of these sectors, definitely it would be a great asset for us to understand how their process flows or how the portfolios are. So definitely it's a great to work with all these kind of best, which you got a chance to work with these clients. And have you won project, you can tell a little bit, you're really proud of. Yeah, so when I'm working with a craft domain, so where I got a chance to work with, so where customers are looking for. Like where they are exactly looking for how. What kind of invoice it is basically more of their report into invoice where it got paid and how many are due. So they are focusing mainly on old use. So there are many more use. So very less minute compared to the paid ones. So that's where I found it's very interesting report. So there are certain challenges where we are unable to come up with a certain using key, that's functional is like, for example, if I take where we are trying to showcase to cumulative sales on a month on month for last 12 months from today. So that's where it's a big challenging. So at that moment, where the function called is on our. There is a fund that's on called is on. Is in scope, that's function exactly match it for the requirement, the values that we are looking for. So that's where I found very interesting like took more time. So which that's function which that's function when we are looking for. So that's where it's quite challenging and with respect to the customers as well, they're looking for. The inside like where it gives more over use. So which aircrafts which company of aircrafts are giving more over use. So that's where they are looking very excited for the from the report level. So that's where I felt very interesting report. And yeah, we have, I think when we build reports and doing the data data engineer stuff, we have the one part is it's the development parts. The other part is the designing part from the dashboards, but there is another I think for my perspective, the hardest part, it's the stakeholder management. How do you handle the stakeholder requirements from your perspective? Yeah, me. Yeah, me. So we need to understand their requirements. We need to connect on a daily connects. We need to ask for the feedback. So we need to check how. What kind of is the insights they are getting from the report or not? So we need to better understand from their conversations. And we should, if we can able to come up by having certain conversations, we can able to understand like what they're looking actually for. So what kind of objectives they are looking forward to showcase to their leadership teams. So that's how we need to ensure that we are coming up with the right insights. So that's where I feel it's good to have the discussions with the stakeholders on a frequent connects. So that's where it gives us good perspective, understand what exactly they are looking for. And yeah, for beginners, what team tips can you give? We have all these Microsoft certifications on the Microsoft learn. You have PL 300 and the DPS 6. How important do you think are these certificates actually? Yes, me. So if you ask me like most companies are preferring for who are having the certifications, right. So we will be thinking that it might not be much interesting or good to take, but most of the learners. But I would suggest to take the exams and try to initially go with the Microsoft learn. So where, for example, if you want to take PL 300 exam, I would suggest to go and type in the Google PL 300 Microsoft learn. So where you'll find a page and where you'll be having certain modules. So where there will be having four modules in each module, a topic wise, you will be having learning paths. So if you can complete the learning path, then you can able to start going on to the power be application tool and you start practicing the hands on. So that's where you can able to answer the when you go for the PL 300 exam. So also, you would suggest to join the data days. So currently, fabric data is running and to be ending by August 6 in upcoming session in upcoming future. If you can join the Microsoft's fabric data days or whatever the sessions they are conducting so that you can able to avail the free authors, you can register. So where you can get a chance to get the free or just and you can start up playing for the exam. So that's where I would suggest to start. And I see on your profile on LinkedIn, you are a fabric community super user. What is this for the eye? Yeah, so basically this recognition I got where I used to provide the solutions to the questions that are getting posted around the world in the fabric community. So it can be related to desktop or be disturbed. It can be related to power be, it can be related to power be for query. It can be related to power be workspace service level. So where I used to start answering the questions that are whichever I'm aware of. So there are certain kind of where following they will be having like where we need to get certain solutions has accepted. So that's where it brings us the badge of a big super user. And okay, let's think. I'm such a little seed is and it says awesome. When it comes to you and say, hey, you get all all the money and, and yeah, all the power of people and so on developer, you will have what feature will you develop? What feature? I did not get your question, can you repeat what? Yeah, when you can say to Microsoft, they say, give you all all the resources you need. What feature for for a break or call be I will you develop? Okay, feature for my side, I would. I never started off thinking of this idea. I think you're on this part, Mirko. Yeah, I have an all all sessions, a lightning or rapid fire around. I asked questions and you give a give a short answer. So coffee tea or energy drink during development energy drink. Dark mode or light mode. Sorry, dark mode or light mode light mode. Favorite power be I feature. A brick for be a feature. One one any of any of the future. What's the your your favorite? Yeah, data. Legend. Okay, most underrated Microsoft technology. I don't have anything in my mind. One, everyone should know one one. One that's function everyone should know about. Calculate. But one feature you love Microsoft to add on on. On power be I. So as of now I'm finding in the data modeling, when I see the table view, I'm going to like certain columns will be on the right side. So if if if Microsoft can provide a functionality of dragging and placing reordering the columns. So that's that that I have that in my mind. Yeah. Is there one book podcast YouTube channel every data professional should listen to. I would say. I did have many but right now I'm not getting into my mind. What your favorite productivity had it. So my third productivity habit is to daily you can spend some time on your learnings, whichever you feel not. More good in so that's where I feel. To start spending some time on a daily basis so that would improve are definitely our. Yeah. So then my mind one of my closing question is what did you think power be I looks like in three years. So definitely we can see many updates many many updates and also will be seeing how the fabric is getting a world from the current preview features to generally available and working directly reporting in the fabric environment itself. So that's where I want to see how the fabric is evolving. Also, yeah, so then my last question is when when I should invite or the next guest who should be in what question should I ask them. In the next podcast what questions you want to ask. Yeah, I should how should I invite and what questions roll I ask it. Yeah, so mostly you can ask on the fabric part I would suggest to go go with the fabric on kind of lake houses, where houses, so how how the data is. How which kind of how the volume is how they what the real time intelligence and what kind of applications are currently organized are using so that's where you can go with completely. And then actually in the fabric environment and have you an idea for the next guest. I don't have I'm happy to hear a little bit. Yeah, so yeah, this was awesome awesome session. Thank you for staying here with me these hour and yeah, thank you for joining the M65 podcast and sharing your experience. Yeah, building scalable power by the I solutions Microsoft fabric enterprise analytics and yeah, joining all data into real business value. So thank you for staying here with me and yeah, it was amazing sessions. Thank you. Yeah, thank you. Thank you, Michael. Thank you for this. Thank you all for joining in. Bye. [BLANK_AUDIO]