Sumit Saraswat

Yeah. So, I've started my career with MetLife, and I have worked there for approximately 7 to 7.5 years as a business analyst or data analyst. In 2019, I joined Adobe Systems as a senior business owner, where I was part of the customer experience domain. We needed to keep track of the performance of the agents who were resolving customer queries via calls or chats. At Adobe, I created different Power BI dashboards. We used to get data from different sources, such as our SQL Server, Excel, and various JSON and KPIs. We would load those BI datasets into our Power BI system and build visualizations based on the requirements of the stakeholders. Most of the time, it was an agent performance dashboard where we created KPIs, like product 20 percentage, online percentage, and contact per handled per day, contact handled per hour, kind of KPS, which we built on the visualization tool. Apart from that, I've also created the AUX usage report dashboard, which was based entirely on Power BI. Most of the time, we would get queries from different managers and teams, saying, "We need this agent data, so we need to fetch that data from a particular server." We would then create reports based on the requirements. That's it from Adobe. In 2023, I joined Oracle as a senior business analyst. In my role here, my responsibilities include maintaining Power BI dashboards and OAC dashboards. Apart from that, we got requests from stakeholders saying, "You need to prepare some reports for a particular audience and the partner side." So, we would prepare those reports and provide insights. In Oracle, we get requests through Jira. When we explore the Jira request, we observe what all KPIs they need or what all requirements they have. Based on the requirements, we provide the data. I guess that's it from my end.

So it totally depends upon the requirements. So which method we use, it could be direct query or import. So if we want to improve our efficiency of the dashboard or report, we preferred import because all the data saves on your data model. But on the other hand, indirect query, the all the data saves on the server side. So we always prefer indirect query if there's no need of daily refresh of the data or the requirement is not urgent for a refresh. So we prefer to use indirect query. Import data. What it does, it creates a cache memory in Power BI Desktop and stores the data in our model. So the retrieval of the data is very fast compared to the direct query.

Yeah, I have worked on restricting data to role level security. So for that, it depends on the user. If the user belongs to a different region or different country, we provide the RLS based on their criteria. We use DEXs. If a particular user belongs to a specific region, we provide the region equal to that region. And based on the role, we restrict them to use specific data. So, role level security is basically two types. One is static and dynamic. In static, we provide access based on the role, where they belong, and what kind of database team they belong to. In dynamic, we pull their username based on the user principal name or username. And with the help of this, we provide dynamic role level security to the user. Dynamic role level security helps us to restrict thousands of users using our dashboard based on the dynamic rollover security. And we create access based on their email ID. We map with that table and map it into our data model. And based on that, we pass the username to provide dynamic row level security.

Should a particular post testing application security. No, I don't have an answer for that right now. We implement rule level security or object level security, but didn't understand this question completely.

What are the key considerations we would recommend? So most of the time, if we're working on a complex text, we make sure it's concise. Variables should be there, and we don't use indexes that take most of the memory. So, we check with the performance visualizer if there are dexes that are taking up memory of the cache. And if they are taking too much memory, we try to find those dexes that are taking less memory compared to the others.

When we're slow running complex text, we determine if it should be optimized for rewriting if our current debt, which we have created, is created, taking too much time to execute, and we check the performance of the particular decks. If it is good or not good, taking too much time in a performance analyzer. So we try to replace that with better decks.

Customers contribute. Yes. So, in here, we are using a wildcard character to search for a specific country. So, what can we do? We replace that with equals Germany. This will help us to find the data where the country customer.country is Germany. Or, second thing, if needed, we can, as per the requirement, fetch only those columns instead of all or instead of asterisk. We only need to fetch the columns that are required in our report, such as product ID, customer ID, revenue, whatever it is.

I don't have answer for that. Oh, let me try, I guess.

Reporting is so isn't that accurately reflect complex. I don't have an answer for that either.