How to Compress a CSV File (Reduce File Size)

Method 1: Lossless Compression (ZIP / GZIP)

The fastest way to make a CSV smaller without changing a single value is to run it through DEFLATE/ZIP or GZIP. CSV is highly compressible because it is repetitive plain text: the same delimiters, the same category strings, similar numbers row after row. A 100 MB export can routinely become a 5-15 MB archive.

• Use it when: you need to email, upload, or store the file and the data must stay exactly as-is.
• Tradeoff: the recipient has to extract the archive before opening the CSV, and the file is the same size again once expanded.

Method 2: Drop What You Do Not Need

Most oversized CSVs carry columns nobody downstream uses and rows outside the range that matters. Removing them is usually the largest single reduction, and unlike compression it makes the file smaller for every tool, not just in transit.

• Keep only the columns your report or import actually reads.
• Filter rows to the date range, region, or segment you work with.
• Split the file if different consumers need different slices (see how to split a CSV file).

Method 3: Remove Duplicate Rows

Exports that union multiple sources or run on overlapping schedules accumulate duplicate rows that inflate size and distort any aggregation. Deduplicate on the columns that define a unique record, exact-match for identical rows, fuzzy matching for near-duplicates with formatting differences. See how to remove duplicate rows from a CSV file.

Method 4: Normalize Oversized Values

• Round floats: 3.1415926535 stored as 3.14 across a million rows adds up.
• Shorten timestamps: if you never use the time, 2026-01-15T00:00:00.000Z becomes 2026-01-15.
• Trim whitespace and collapse verbose repeated strings into short codes where a lookup exists.

Method 5: Switch to a Columnar Format (Parquet)

For data you keep or analyze repeatedly rather than hand to a colleague, Parquet stores each column together and compresses it independently, often producing a file a fraction of the CSV size that is also faster to query. The tradeoff is that it is a binary format, not something you can open in a text editor. See CSV vs Parquet for when the switch is worth it.

Which Method Should You Use?

• Emailing a file once: ZIP it (Method 1).
• A tool or database keeps timing out: drop columns and rows, then deduplicate (Methods 2-3).
• Recurring export that is always too big: fix it at the source query, then normalize values (Method 4).
• Long-term storage or analytics: convert to Parquet (Method 5).

Checklist Before You Delete the Original

• Row and column counts match what you expect after each reduction step
• The compressed archive extracts cleanly and the CSV inside opens normally
• Spot-checked IDs, dates, and totals against the source file
• The recipient can open the format you sent (ZIP vs Parquet)

Frequently Asked Questions

Doing This in How To CSV

How To CSV covers each step as a browser-based tool, nothing is uploaded to a server:

• Compress CSV/XLSX for lossless ZIP compression (Method 1)
• Cleaning and column selection for dropping unused fields and trimming values (Methods 2 and 4)
• Fuzzy Dedupe and Logical Dedupe for duplicate removal (Method 3)
• CSV to Parquet for the columnar-format switch (Method 5)