Publishing Your Research · Lesson 2 of 5
Organizing a Research Repository
~12 min
The Concept
You spend an afternoon reorganizing the nitrate project into a structure a stranger could actually navigate, separating raw data from processed data, numbering the analysis scripts by the order they run in, and writing a top-level README that explains what's where.
A well-organized research repository typically separates raw data (never modified after collection) from processed data (the output of documented cleaning steps), keeps analysis code numbered or otherwise ordered to show the pipeline sequence, and includes a README (a short welcome file that orients a new visitor) that orients a stranger in under two minutes.
The Analogy
A well-organized project folder is like a well-labeled toolbox, with wrenches in one drawer, screwdrivers in another, and a card taped to the lid explaining what each drawer holds. A messy toolbox with everything dumped in one pile might still contain the same tools, but good luck finding the right one in a hurry.
Code (optional)
bluewater-nitrate-analysis/
├── README.md # what this project is, how to run it
├── data/
│ ├── raw/ # never edited after collection
│ │ └── gw14_nitrate_2020_2025.csv
│ └── processed/ # output of scripts/01_clean_data.py
│ └── gw14_nitrate_clean.csv
├── scripts/
│ ├── 01_clean_data.py
│ ├── 02_exploratory_analysis.py
│ ├── 03_hypothesis_test.py
│ └── 04_generate_figures.py
├── figures/
│ └── nitrate_trend_2025.png
├── requirements.txt # exact package versions used
└── LICENSELine by line
- The raw/processed split protects the original data from ever being silently altered.
- Numbered scripts make the pipeline order self-documenting, anyone can tell 01 runs before 04.
- requirements.txt pins exact package versions, since a different pandas or numpy version can subtly change results.
Why Real Researchers Care
This structure closely follows widely adopted conventions like Cookiecutter Data Science, used across academic and industry research teams specifically because ad hoc folder structures are one of the most common, preventable causes of irreproducible work.
Quick Check
Q1. Why should raw data live in a folder that's never edited after collection?
Your Goal
Sketch a folder structure for your own Bluewater Basin capstone project (Lesson 5 of this mission), following the raw/processed/scripts/figures pattern above.
Hint: You don't need every folder to have contents yet, the structure itself is the deliverable for this challenge.
Teach It Back
Explain to a teammate why numbering your analysis scripts (01_, 02_, 03_) matters for reproducibility.