Analytics2023—2024
Data Analytics & Science Monorepo
Eight end-to-end analytics projects — Python, SQL, Tableau, and Power BI — in one auditable archive.
- Python, SQL, Tableau, Power BI
- Reproducible, auditable analysis
- Exec decks and engineer notebooks
- Projects
- 8 end-to-end
- Toolchain
- Python · SQL · Tableau · Power BI
- Notebooks
- ~37 MB of analysis
- Contract
- data + code + deliverable
01
Overview
A monorepo collecting eight complete data projects across the full analytics toolchain: Python notebooks for EDA and modeling (IPL, Pokemon, Pima Diabetes, Game of Thrones, Automobile, CardioGoodFitness), and business-intelligence builds in Tableau and Power BI (F&B Sales, Industrial Combustion Energy Use) with their SQL KPI layers, source spreadsheets, and stakeholder decks. Every project ships its dataset, its notebook or workbook, and a rendered PDF — the whole chain of evidence in one place.
02
Problem
Analytics work scatters: a notebook here, a Tableau workbook on a laptop, a KPI query in an email. Six months later the insight survives but the evidence does not. This repo is the fix — one place where every project keeps its data, its code, and its final deliverable together, so any result can be re-run and any chart can be traced to its query.
03
Goals
- Archive each project as data + analysis + deliverable, never just the output
- Cover the real analyst toolchain — Python for depth, SQL for KPIs, Tableau and Power BI for stakeholders
- Make every result reproducible from the committed CSVs
04
Solution
Each project lives in its own folder with a consistent contract: the raw dataset (CSV or Excel), the working analysis (.ipynb, .twbx, .pbix, or .sql), and a rendered artifact (PDF or PPTX) for reading without tooling.
The BI projects go beyond dashboards: Industrial Combustion Energy Use includes the SQL KPI definitions, twin Tableau workbooks, a Power BI dashboard, and the presentation actually used to communicate findings — the full path from warehouse query to executive slide.
Tooling
Tech stack
- Python
- Jupyter
- SQL
- Tableau
- Power BI
- Excel
- pandas
- scikit-learn
Friction
Challenges
One repo, three toolchains
Python notebooks, Tableau workbooks, and Power BI files have nothing in common except the data. The folder contract (data + analysis + rendered output) is what keeps the repo navigable across formats.
Large binary analysis files in git
Notebooks with embedded plots and BI workbooks grow fast — the repo carries ~37 MB of notebooks alone. Committing rendered PDFs alongside means reviewers never need to execute anything to evaluate the work.
Takeaways
Lessons learned
- A consistent per-project contract beats any folder taxonomy — predictability is what makes an archive usable
- Shipping the PDF next to the notebook respects the reader who will never install Jupyter
- BI tools and notebooks answer different questions; keeping both for the same dataset (Automobile appears in both forms) shows the trade-off clearly
Roadmap
Future improvements
- Add README index cards per project with headline findings
- Publish the Tableau workbooks to Tableau Public for one-click viewing
- Migrate shared loading/cleaning code into a small utilities package
Rebuilt for this site from the original repository — source history