Tech Layoffs — SQL Data Analysis (2020–2023)

A data analysis project on global tech sector layoffs using MySQL — covering data cleaning and exploratory analysis.

Tool: MySQL Workbench 8.0
Dataset: layoffs.csv — 2,361 rows, 9 columns

---

Files

├── layoffs.csv                       # raw dataset
├── DataCleaning.sql                  # cleaning pipeline
├── ExploratoryDataAnalysis.sql       # EDA queries
├── tech_layoffs_dashboard.html       # interactive dashboard
└── README.md

---

Dataset

Column	Type	Description
company	TEXT	company name
location	TEXT	city
industry	TEXT	sector
total_laid_off	INT	headcount cut
percentage_laid_off	TEXT	fraction of workforce; 1.0 = full shutdown
date	TEXT → DATE	announcement date
stage	TEXT	funding stage at time of layoff
country	TEXT	country of HQ
funds_raised_millions	INT	total capital raised

---

Data Cleaning

The raw table is never touched. Everything runs on a staging table (layoffs_staging), and a second one (layoffs_staging2) is created to handle the deduplication step since MySQL doesn't allow deleting directly from a CTE.

Steps followed:

1. Remove duplicates — used row_number() partitioned over all columns to flag duplicates, then deleted rows where row_num > 1 from layoffs_staging2.

2. Standardize data — trimmed whitespace from company names, collapsed crypto industry variants ('Crypto Currency', 'CryptoCurrency' etc.) into 'Crypto', removed a trailing period from 'United States.', and converted the date column from text to a proper DATE type using str_to_date().

3. Handle nulls — blanks in the industry column were set to null first, then filled in using a self-join where another row for the same company had the value. Rows where both total_laid_off and percentage_laid_off were null got deleted since they're not useful for analysis.

4. Drop helper column — row_num was only needed for deduplication so it gets dropped at the end.

---

Exploratory Analysis

Queries are in ExploratoryDataAnalysis.sql and cover:

• max layoffs and companies that fully shut down (percentage_laid_off = 1)
• totals by company, industry, country, year, and funding stage
• monthly breakdown and rolling cumulative total using a window function
• top 5 companies per year using a two-CTE approach with dense_rank()

dense_rank() is used instead of rank() so ties don't create gaps that would push companies out of the top 5 filter.

---

What the data shows

• 2022 was the worst year — 161,711 layoffs, compared to just 15,823 in 2021. The 2021 low came from zero-interest-rate hiring booms; 2022 was the correction.
• January 2023 was the worst single month — 84,714 layoffs. Google, Microsoft, Amazon, and Salesforce all cut in the same month.
• Post-IPO companies accounted for the most layoffs — public market pressure to cut costs hit harder than early-stage companies.
• US made up ~66% of all layoffs — 258,159 out of 386,379. India was second at 35,993.
• Consumer and Retail were hit hardest by sector — 46,682 and 43,613 respectively.
• 116 companies laid off 100% of staff — some had raised hundreds of millions before shutting down.

---

How to run

1. Import layoffs.csv into MySQL as a table called layoffs
2. Run DataCleaning.sql — this creates and populates layoffs_staging2
3. Run ExploratoryDataAnalysis.sql against layoffs_staging2
4. Open tech_layoffs_dashboard.html in a browser for the visual summary

---

MySQL Workbench 8.0 · dataset from public layoff tracking records