Chapter 2 — Tidy and the Cleaning-Log

Tidy was a small raccoon-tween. Her fur was a mix of warm gray, cream, and soft black. The markings on her face looked like a friendly, rounded mask, never sinister. Her hands were quick and gentle, always busy. A small, pale-gray cloth notebook, her cleaning-log, was always strapped to her side. It had “CLEANING LOG” written on the cover in neat block letters. When she worked with data, the notebook lay open on her workbench.

This notebook was more than just paper. It was Tidy’s craft, her conscience, her way of working. She wrote down every decision she made about the data. First, she read Catch’s collection notes. Who gathered the data? What was it about? Why and when was it collected? Only then would Tidy open her cleaning-log. She recorded what cleaning choices she made and why. She noted what the data looked like before, and what it looked like after.

Without this log, any analysis built on the data would hide invisible decisions. With it, every choice could be inspected and questioned. This was vital. Tidy taught the data-cleaning primitive. This was the skill of preparing data without pretending the preparation was neutral.

Real datasets were always messy. Imagine a list of everyone’s favorite ice cream flavors. What if someone wrote “choclate” instead of “chocolate”? Or listed “vanilla” twice? Or what if a whole row was just blank? That’s what real data often looked like: missing values, duplicate rows, typos, strange numbers, or mixed-up formats. Cleaning was necessary to make sense of it all.

But every cleaning choice changed the dataset’s meaning. If you dropped rows with missing values, you might accidentally favor only complete records. If you filled in missing numbers with an average, you smoothed over real differences. Removing unusual numbers, called outliers, might take away the most important data points. Renaming categories could sometimes make you lose information. The skill wasn’t avoiding cleaning. That was impossible. The skill was making every cleaning choice visible in the cleaning-log.

Tidy never called it “data housekeeping.” She found that term annoying. It sounded like dusting shelves. But data cleaning was never just tidying up. It was making big, important choices. “Cleaning is not neutral,” she would say. Her voice was firm. “Every cleaning choice changes the meaning. Document the choices. The next analyst — or future-you — needs to know what you did. They need to know why, and what the data looked like before.” Without the cleaning-log, the analysis couldn’t be trusted or repeated. Tidy saw cleaning as a first-class analytical job, not just a chore before the real work began.

Tidy often thought back to her childhood. Her family had been the grain-sorters for their small village. They separated the annual harvest into different piles: kitchen-grade, mill-grade, and seed-grade. It wasn’t just pouring grain into bins. It required careful, clear choices about what counted as which grade. The sorter who couldn’t explain her grading was the sorter the millers stopped trusting. By age six, Tidy had learned that sorting was a choice. And those choices had to be visible to be trusted.

When she walked to the DataForge academy at twenty-two, Datum had asked her one question. “What is data cleaning?”

Tidy had answered without hesitation. “It is preparation-with-integrity. Every cleaning choice changes the meaning. Document the choices. The cleaning-log is the conscience of the pipeline. Without it, the analysis is built on invisible decisions.”

Datum had simply nodded. “You are appointed.”

In her workshop, Tidy began every first-day lesson the same way. The room smelled faintly of old paper and new data. Students, a mix of curious and nervous, watched her. She opened her cleaning-log on the workbench. The pages were crisp and white. She wrote the dataset name at the top of a fresh page.

“I am Tidy,” she announced. Her voice was clear. “The data-pipeline primitive I teach is cleaning. The move is document every choice. Every cleaning step has alternatives. The choice between alternatives shapes the analysis. Make the choices visible.”

She then explained her “cleaning scaffolds,” a set of guiding steps. She pointed to a diagram on the wall.

“First,” she said, “read Catch’s collection notes. Cleaning depends on knowing how the data was collected. Was it from a survey? A sensor? A lab experiment?”

“Next, inspect the data before cleaning. Look at the first twenty rows. What do you see? Check the summary statistics. What’s the average? The highest? The lowest? Look at how each variable is spread out. Know what you’re starting with.”

“Then, identify the cleaning issues. Are there missing values? Duplicate rows? Typos? Outliers – those really strange numbers? Mismatched formats?” She paused, letting the words sink in.

“For each issue, list the alternatives. For missing values, you could drop the row. Or fill it in with the average. Or the median. Or even a prediction. Or just leave it missing. Each alternative has trade-offs. No choice is perfect.”

“Choose deliberately,” Tidy stressed. “Don’t just pick the first option. Choose with awareness of what your choice means.”

“And finally, document the choice in the cleaning-log. Write the date. The dataset name. What you did. Why you did it. What the data looked like before, and what it looks like after.” She tapped the open log. “This is your proof. This is your honesty.”

She added two more critical rules. “Always preserve the original data. Never overwrite it. Always work on a copy. And make the cleaning-log available to anyone who uses your data. That includes future-you. The log is part of the dataset.”

“I sometimes make a cleaning choice that I later realize was wrong,” Tidy admitted. Her gaze swept over the students. “That’s not failure. That’s why the log exists. I can revisit, change the choice, and update the log. The transparency is the practice.”

When students asked Tidy whether data cleaning was hard, Tidy always gave the same answer.

“It is not hard,” she said. “It is deliberate choosing, plus careful documenting. Every cleaning choice changes the meaning. Document the choices.”

She closed her cleaning-log gently. The next dataset waited to be cleaned.

---