yotam22/watches

🕰️ Exploratory Data Analysis of Luxury Watch Prices

Overview

This project analyzes a large dataset of luxury watches to understand which factors influence price.
We focus on brand, movement type, case material, size, gender, and production year.
All work was done in Python (Pandas, NumPy, Matplotlib/Seaborn) on Google Colab.

---

---

Dataset

• Rows: ~172,000
• Columns: 14
• Unit of observation: one watch listing

Main columns

• name – watch/listing title
• price – listed price (string with symbols)
• brand – brand (e.g., Rolex, Omega, Patek Philippe)
• model – model / reference
• ref – reference number
• mvmt – movement (Automatic / Quartz / Manual / etc.)
• casem – case material (Steel, Gold, Titanium, …)
• bracem – bracelet material
• yop – year of production (may include text like “Approximation 1998” or “Unknown”)
• cond / condition – condition label
• sex – Men / Women / Unisex
• size – case size (e.g., “40 mm”, “42 x 50 mm”)

---

Engineered Features

• price_num – numeric price extracted from price
• size_mm – numeric size (mm) extracted from size (first numeric token)
• year – four-digit year extracted from yop

---

Research Question

What are the main factors that influence the price of luxury watches?

Sub-questions

1. Do certain brands have significantly higher prices than others?
2. Does movement type (Automatic / Quartz / Manual) affect price?
3. Does case material (gold / steel / titanium / etc.) influence price?
4. Are men’s watches more expensive than women’s?
5. Is there a relationship between watch size (diameter) and price?
6. (Optional) Does production year impact price?

---

Data Cleaning

1. Removed duplicates (df.drop_duplicates()).
2. Price → price_num: stripped $ and ,, converted to float.
3. Size → size_mm: parsed first number (e.g., “42 x 50 mm” → 42); set values > 60 mm to NaN.
4. Year → year: extracted first 4 digits; set "Unknown" and years < 1900 to NaN.
5. Basic missing-value checks and consistency validation.

---

EDA Highlights

1) Price Distribution

• Histogram of price_num (limited to ≤ $50,000) shows a right-skewed distribution.
• Most watches are under $10,000, with a small number of ultra-luxury outliers.

2) Price by Brand (Top 10)

• Boxplot across top-frequency brands.
• Patek Philippe, Rolex, Audemars Piguet show the highest medians and ranges.

3) Price by Movement Type

• Boxplot by mvmt.
• Automatic/Manual watches are significantly pricier than Quartz.

4) Price vs Size

• Scatter of size_mm vs price_num.
• No strong linear relationship; size is a weak predictor of price.

5) Price vs Year (Optional)

• Scatter and correlation of year vs price_num.
• No clear trend: both vintage and modern pieces can be expensive.

---

Insights & Answers

• Brand: Strongest driver of price; prestige and rarity dominate the high end.
• Movement: Mechanical complexity (Automatic/Manual) correlates with higher prices.
• Case Material: Precious/high-end materials tend to increase price (magnitude varies by brand/model).
• Gender: Men’s watches are slightly more expensive on average (brand-dependent).
• Size: No clear relationship with price.
• Year: Weak/none overall; vintage vs. modern effects are brand/model specific.

---

Conclusions

The luxury watch market is brand-driven. Prices are primarily shaped by brand reputation and mechanical craftsmanship (movement), with materials providing additional lift. Size and age are poor standalone predictors.

Bottom line: Brand + Movement (craftsmanship) > Material >> Size/Year.

---

Reproducibility

1. Open the notebook: Copy of Assignment #1 - EDA & Dataset.ipynb.
2. Upload Watches 2.csv to the runtime.
3. Run cleaning cells to create price_num, size_mm, and year.
4. Execute the EDA cells to regenerate all visuals and summaries.

Environment: Python (Pandas, NumPy, Matplotlib, Seaborn) on Google Colab.

---

Project Info

Author: Yotam Gil
Program: Reichman University — Economics & Data Analytics
Year: 2025

---

loom video

https://www.loom.com/share/07986da47c5645fdbf06a97defd987f1