🏙️ Louisville Payroll Analytics – Part 5
🔁 Looking Back at the Project
From the outside, this repository now looks clean:
SQL scripts, views, a Tableau dashboard, and a README.
From the inside, it was a month of small battles:
- PostgreSQL permissions and
pg_hba.confheadaches - DBeaver import failures and mis‑mapped columns
- Wrong joins, missing rows, and totals that made no sense the first time
- Iterating structure and naming until things finally felt right
And that’s exactly why this project matters to me. It’s not just “a finished dashboard”, it’s the proof that I can hold through the boring and frustrating parts until I get to insight.
🧨 Biggest Surprises
Environment problems took longer than SQL logic.
Setting up roles, granting schema privileges, and getting DBeaver to cooperate cost more time than writing the core queries.Simple checks saved me.
Row counts before and after joins, sum of payroll by year, basic sanity checks—these caught most of my early mistakes.Views are underrated.
Moving logic into views made Tableau simpler and made the overall architecture feel more like a real BI stack.
💡 What I Learned (Technically)
- How to design a fact table (
salary_data) that is friendly for analytics. - How to use indexes to keep queries fast on real data.
- How to build window‑function‑based views (rankings, running totals, variance analysis).
- How to structure a Tableau dashboard around a narrative, not just charts.
- How to keep everything version‑controlled and reproducible via GitHub.
❤️ What I Learned
This project thaught me two things:
More difficult than expected Satisfaction hits when you finally run a complex query, refresh the dashboard, and the numbers make sense.
📊 Tableau Visual: Links
The dashboard shows the overall budget by department, overtime per department, salary ranking per job title, and the yearly trend comparing number of employees and average overtime rate.
🎯 What’s next now?
General review about what I made so far and my future projects.
⚡ Credits
© 2025 Pietro Di Leo. From Operations to Data, one Commit at a Time.