Ankit Singh

Ankit Singh. I have 8 years of experience in ecommerce and insurance domains. The recent one is in ecommerce, in service analytics, where I build dashboards on Tableau and Power BI for multiple US clients. And so the role involved all the things from requirement gathering till delivery. We had kickoff calls with clients, business workshops to discuss all the requirements and get a sign off on those. Then we discussed the dashboard design as well. We built the data pipelines and then used these data pipelines once built in the Tableau or Power BI dashboard as per the customer's requirement. The warehouse that we used was in Google BigQuery, which is similar to SQL. Other than that, we have used dbt cloud for the ETL process. So that's pretty much from sales analytics. The remaining 5 years of experience I have is in SSP India Private Limited, where this was an insurance domain. We were tracking the customer's policies and renewals or insurance, etcetera. I've worked on some Tableau dashboards. Other than this, the remaining 2 years of experience I had is in Tata Consultative Services. Again, there were some changes that were assigned to us for multiple clients in the Tableau dashboard. So that's pretty much what I have.

Implement a test plan to validate the accuracy of an ETL process. Okay, so the test plan would be for Amazon. There are several scenarios to consider. Let's say if you are validating Amazon sales data for a particular customer. So what we do is download the reports directly from the client's platform or from Amazon, and then manually calculate the numbers in Excel. And then we check the final table in our warehouse to see whether the numbers match. If they do not match, then there might be a mistake in the calculation or a coding error that we introduced, so we need to fix that. This is one of the test plans that we use to validate the accuracy of an ETL process.

And given a complex dataset, what are the steps you take to prepare it for the use in a Power BI dashboard? So the first step that we take is to check for unacceptable values. Like, if we are expecting a string value, whether we are getting nominal values or null values in there. If we are expecting integer values, whether we are getting only integer values there or there are some string values also present. So we need to check all these. 1st, we need to check all the columns for the types of values that we are getting. So we do a distinct value check and the range of values, like how much we are getting, whether there are outliers or not, null value checks. Then, we do the exploratory data analysis. After that, we transform it according to our needs, and then we can use it in a Power BI dashboard.

We coordinate with stakeholders to convert business requirements into technical specifications, particularly focused on data needs. So we conduct business workshops. We discuss what are their data needs, and then we create the business requirement document, then they sign off. And so, basically, what we want to know from the business stakeholders is what the final table or presentation layer table structure would be. And then we transform the raw data into a final table format, and then we use it in the Power BI dashboard. So that's pretty much what our focus is on.

Project coordination with hands-on technical work, such as writing SQL queries and performing QA activities. These things go hand in hand. Project coordination involves setting up meetings, like daily huddles with the team and weekly huddles with the customer to share progress. Writing SQL queries and performing QA activities happen every day, going hand in hand.

Performance problem in an SSRS report. Would you identify and resolve the problem? I haven't worked on this as an SSRS report, but if I'm facing any performance problem, let's say in one of the Tableau dashboards. So I usually check the Internet how to resolve these and then identify that Tableau has a feature to record the performances. So I recorded it and saw that there are some performance issues. If we do a manual sort, it takes more time. That way, we can assess the performance issues. Similarly, we just need to track the time, which processes are taking more time. That way, we can work on those and resolve the issue.

Input correct username into username field. Input correct password into password field. Click login button. Check if user is redirected to the dashboard page. So, let's verify successful login. Input correct username into username field. Input correct password into password field. Click login button. Check if user is redirected to the dashboard page. So, like, shortcomings in the testing logic. So we can check for incorrectness as well. Let's say, we just click on login button and see if the user is directed to the dashboard. Input correct password into password field. Input correct username into username field. Click login button. Check if user is redirected to the dashboard page. We can enter incorrect password in the password field, correct username in the username field, and then click log in to see if the user is directed. We can input incorrect username and correct password. We can keep password field blank. We can keep username field blank. So there are multiple use cases that we can test to see if the login is successful or not.

Select ID, first name, last name from users where is active equal to 1 and (created date is greater than '2023-01-01' or updated date is less than '2023-01-01'). Logical mistake: The condition "created date is greater than this" should be "created date is greater than '2023-01-01'". The condition "updated date is less than this" should be "updated date is less than '2023-01-01'". Impact: The query will return users who are active and have a created date after January 1st, 2023, or an updated date before January 1st, 2023. However, the condition "created date cannot be greater than updated date" is not a logical mistake in this query, it's actually a correct observation. The intention of the user seems to be to filter users based on a date range, but the query is currently set up to return users who fall within the range of '2023-01-01' or before. If the user wants to see data for 1 year between 2022 and 2023, the condition can be changed to "created date is greater than '2022-01-01' and updated date is less than '2023-01-01'".

How did you approach the design and execution of a full test plan for a large-scale web application project? Okay, full test plan. So, basically, I would use a top-down approach. Firstly, I would define the high-level targets, such as what the web application should be about. Then, I would break down the application into smaller functionalities. For each functionality, I would create test cases. Once those test cases are passed, it can be said that the web application is ready.

Power via visualization techniques to convey complex data patterns to nontechnical stakeholders. So, basically, there are multiple line charts, bar charts, and various types of graphs that we can show to the customer or business stakeholders to show the different data patterns, like let's say, where there are spikes in the line chart. We can explain to the business stakeholders why that is. This way, we can convey the different data patterns to nontechnical stakeholders.

Of how you used ETL tools to enhance data analysis and reporting in your previous projects. In my previous company, we used dbt cloud. So we added null checks as part of adding a layer of checks so that the data is not corrupt in the initial state and gives us a better clarity there. That's what we implemented in our previous project.