Blog

The biggest surprise wasn't the tools — Python and SQL took weeks to feel natural, not months. It was realizing how much of audit actually is data analytics. Reconciling a bank statement against SAP and reconciling a messy CSV against a source-of-truth table are the same instinct wearing different clothes: find where the numbers disagree, and figure out why.

What's been harder is the part nobody warns you about — sitting with a half-finished dashboard and not knowing if the chart you picked actually answers the question someone asked. In audit, the question is usually fixed. In analytics, half the job is figuring out what the real question even is before touching the data.

If I could tell my past self one thing: clean data discipline carries over completely. Confidence in deciding what to build doesn't — that only comes from building a lot of dashboards, including the wrong ones.

Coming from Excel, SQL felt like learning to talk to a very literal, very fast colleague. A handful of patterns cover most of what I actually need on real business questions:

Aggregation with GROUP BY — almost every HR or sales question starts here: totals, averages, counts, by department or region.

JOINs across relational tables — once data lives in proper tables instead of one giant spreadsheet, joins are how you put the story back together.

CASE WHEN for categorization — turning raw numbers into bands (e.g. attendance ranges, performance tiers) is something I used constantly building the WorkforceIQ queries.

Window functions for running comparisons — month-over-month or rank-within-group questions, which a single GROUP BY can't answer alone.

Date filtering — sounds basic, but getting fiscal-year and reporting-period logic right is exactly the kind of precision audit trained into me.

I built E-Grocery Insights BD in Tableau and RetailNexus in Power BI — not because I planned a comparison, but because each project called for what was on hand at the time. In hindsight, it turned into a useful side-by-side.

Tableau felt more forgiving for exploratory work — dragging fields around to see what a chart wants to be before committing to a layout. Power BI felt stronger once the structure was already clear, especially with DAX measures sitting close to the data model instead of off to the side.

Neither tool replaced the actual hard part, which is upstream: if the underlying table is wrong, no amount of tool preference fixes the dashboard. That part stayed identical in both projects.