I find out why the obvious answer is usually wrong.
Data Analyst in Toronto — SQL, Python, Power BI, and an accounting background that makes margin and cost problems feel like home turf.
Every project below started with an assumption that turned out to be false.
"Margin is dropping — must be all the discounting." It wasn't. The real cause was something nobody was watching.
"Subscribers churn because they stopped using the product." They didn't. The actual driver had almost nothing to do with engagement.
I'd rather spend an extra day chasing down why a number looks wrong than ship a clean story that isn't true. That's the whole method: start with a business question, follow the evidence instead of the hunch, and stop only when the recommendation is sharp enough to act on.
30,000 real KKBox subscribers. A Python/scikit-learn model that predicts who's about to leave, SQL that quantifies exactly what that's worth, and a dashboard shipped twice — once in Power BI, once as a live web app — because a static screenshot shouldn't be the only way to see the work. The find: $75,964 in recoverable revenue, and churn that tracks subscription mechanics, not listening habits.
8,399 real transactions, one uncomfortable question: where did the margin go? SQL traced it back to its root cause — not the one everyone assumed. See it live. The find: $1.1M in hidden losses, and a story the raw numbers don't tell until you know where to look.
A Business Analyst project and a Pricing Analyst project are next — same method, new questions.
SQL (PostgreSQL) Power BI Python Tableau Excel React
Accounting-trained (B.Com), then a dual postgrad in Data Analytics — which is the long way of explaining why cost structure and profitability questions feel less like a challenge and more like a puzzle I already know the shape of. Currently a Technical & Customer Experience Analyst at Transcom, elbow-deep in customer data and root-cause investigation every day.
Job-hunting for Data Analyst / Business Analyst / Pricing Analyst roles, and shipping one honest, real-data project at a time — limitations disclosed, links live, nothing overclaimed.
📍 Toronto, ON · LinkedIn · shubham.s.sarje@gmail.com