Data Analyst · Jaipur, India
Data Analyst focused on turning messy data into clear decisions
I work with Excel, SQL, Power BI, Tableau and Python. I clean the data, model it, and build dashboards that answer the question a manager actually asked. Currently at Zeetron Networks Pvt. Ltd., Jaipur.
I'm a BCA graduate from the Shekhawati Institute of Engineering and Technology (SIET), now pursuing an MCA at the University of Rajasthan. I have worked with real datasets at Grras Solution Pvt. Ltd., IBV Technologies Pvt. Ltd. and now Zeetron Networks Pvt. Ltd., all in Jaipur.
My work sits between the raw file and the decision. I clean the data in Excel, Power Query or Pandas, write SQL to pull what matters, and build dashboards in Power BI and Tableau that people can read in a minute.
Alongside my job I built the Practice Hub on this page: 170 days of free roadmaps for anyone learning data analytics, with a task, a practice file and a YouTube lesson for every day.
End-to-end analytics work, from the raw file to something a manager can use.
Revenue, RevPAR, ADR and Occupancy across properties, built in Power BI
Regression model estimating used-car prices from market data
300K+ events: funnel, cohort and RFM analysis on 88K users
The tools I use every week. Bars show my own rating and I keep raising them.
Three roles in Jaipur, each one closer to real business data.
Zeetron Networks Pvt. Ltd., Jaipur
IBV Technologies Pvt. Ltd., Jaipur
Grras Solution Pvt. Ltd., Jaipur
Short guides on the methods I use and the things I am learning.
The order I clean any file in, so I never trust a number before I have checked where it came from.
Cleaning in a random order means fixing the same problem twice. A fixed order also leaves a record of what you changed, which is the first thing a manager asks when a number looks different from last week.
Never overwrite the raw file. Every fix should be something you can replay.
Check row count, column names, data types and a sample of 20 rows. In Pandas that is df.info() and df.describe(). In Power Query, switch on Column profile.
One header row, one value per cell, no merged cells and no totals mixed into the data. Set data types early. Dates stored as text are the most common silent bug.
Trim spaces, fix case and unify spellings, so that "jaipur", "JAIPUR" and "Jaipur " become one city. Excel: TRIM, PROPER. Pandas: .str.strip().str.title(). Power Query: Transform, Format, Trim.
Decide for each blank whether to fill it, derive it or leave it. Sometimes it can be derived, for example quantity = sales ÷ price. Check duplicates on the business key such as Order_ID, not on the whole row, then remove them and log how many you removed.
A clean file is not one that looks tidy. It is one where you can explain every change.
Want to practise? Download messy_sales.csv from the Practice Hub and run these five steps on it.
Start from the decision, not the visual. A short framework for KPI-first dashboards.
They show everything the data can show, so the reader has to work out what matters. A good dashboard answers a few questions, quickly.
Ask which decision this page supports. If you cannot name it, you are not ready to build it.
List 3 to 5 questions in the manager's own words: which region is falling behind, which product line drives revenue. Each question becomes one visual.
A number alone means little. Show it against a target, last period or last year. Build these as DAX measures, for example a Total Sales measure and a last-year version using SAMEPERIODLASTYEAR.
Clean in Power Query first: nulls, data types, custom columns. Then build a small model with a date table and clear relationships. Most slow or wrong dashboards are model problems, not visual problems.
KPI cards on top, the trend in the middle, breakdowns below. Keep slicers in the same place on every page and use one or two accent colours.
A dashboard is finished when nothing more can be removed, not when nothing more can be added.
Ranking, running totals and month-over-month change without messy self-joins.
A normal GROUP BY collapses rows into one row per group. A window function keeps every row and adds a calculation across a related set of rows, called the window.
GROUP BY summarises. A window function annotates.
ROW_NUMBER() OVER (PARTITION BY product_line ORDER BY revenue DESC) numbers the rows inside each product line. Filter for 1 to 3 in an outer query or CTE to get the top 3 per group. RANK leaves gaps after ties and DENSE_RANK does not.
SUM(sales) OVER (ORDER BY order_date) gives a cumulative total. Add PARTITION BY region to get one running total per region.
LAG(monthly_sales) OVER (ORDER BY month) brings the previous month's value onto the current row. Month-over-month change is then (sales − previous) ÷ previous.
If you can say the question as "for each group, compare each row to...", you probably need a window function.
This is Days 23 to 25 of the SQL roadmap in the Practice Hub.
The time split behind the Practice Hub and how to stay consistent.
Excel 30 days, SQL 30, Power BI 25, Tableau 15, Python 15, NumPy 10, Pandas 10, Statistics 15 and Machine Learning 20. Excel and SQL get the most time because most analyst work and interviews rely on them, and Power BI comes next because it is the dashboard tool most jobs ask for. Python and ML only go as deep as an analyst needs.
Watching a video is not learning. Finishing the task for the day is.
After Excel, publish a dashboard. After SQL, write a case study that answers 5 business questions. After Pandas, upload a notebook to GitHub. Recruiters open projects, not certificates.
Do not restart. Finish the missed task first, then carry on. Steady progress over 170 days beats a perfect first week.
A four-step method for finding the one step where users drop out, and turning it into a number a manager can act on.
A dashboard that only shows the conversion rate at each step gives you a number with no context. Funnel analysis is about finding where intent breaks down, and for whom.
A funnel does not have a conversion problem. It has a specific problem, at a specific step, for a specific segment. Your job is to find it.
If a user refreshes a page 5 times, that is 5 events but one session. De-duplicate first, or the funnel will look better than it is. In SQL, ROW_NUMBER() OVER (PARTITION BY user_id, session_id, stage ORDER BY event_time) flags the first time each user reaches each stage.
Compare step-to-step conversion, not just the overall rate. Fix the biggest drop first and treat the rest as noise. For example, if most people leave at the product view stage but cart-to-purchase is healthy, the job is to fix discovery, not checkout.
The overall funnel hides stories. Split it by device, new vs returning users, category, traffic source, time of day and price band. A step that looks fine overall can be badly broken for one segment.
Do not say "conversion is low". Say what fixing it is worth. Example numbers only: if 2,000 people reach the cart in a month, 1,000 buy today, and a fix lifts that by 5 percentage points, that is 100 extra orders. At an average order value of ₹1,200 that is ₹1,20,000 a month.
The deliverable of funnel analysis is not a chart. It is a ranked list of specific interventions, each with an estimated revenue impact.
To practise the SQL side, see Days 23 to 29 of the SQL roadmap in the Practice Hub.
Free roadmaps for becoming a data analyst. Every day has one task, a practice file or dataset, and a YouTube lesson. Skills that matter most to analysts get more days, and no roadmap runs past 30.
Take the Challenge: pick Excel, SQL, Power BI or Python, choose your own number of days, submit a file daily and see what is pending.
Get in touch
Looking for data analyst work where I can solve real business problems, not just build dashboards. Happy to talk about full-time roles, freelance dashboards, or the roadmaps in the Practice Hub.
This opens your email app with the message ready to send. If nothing opens, write to anujangid2902@gmail.com directly.