Why zipping sometimes saves 90% and sometimes nothing
You zip a folder of spreadsheets and it drops to a tenth of the size. You zip a folder of photos and it saves four kilobytes. Same tool, same settings, wildly different outcome — and the reason explains most of what people find confusing about ZIP files, including why an archive can be bigger than what went into it. This guide covers what ZIP actually does, what it cannot shrink, what its password option really protects, and when zipping is the wrong answer.
What ZIP does
A ZIP is two things at once: a container that bundles many files and their folder structure into one, and a compressor that shrinks each file individually. The near-universal compression method is Deflate, from 1993, which finds repeated byte sequences and replaces later occurrences with a short reference, then encodes the result with shorter codes for commoner symbols.
The crucial detail is that files are compressed separately, not as one stream. Fifty nearly identical invoices do not benefit from their similarity to each other — each is compressed alone. That is why ZIP loses badly to formats like 7z or tar.gz on large collections of similar files, and why it wins on convenience: any single file can be extracted without touching the rest.
What compresses and what does not
| Content | Typical reduction | Why |
|---|---|---|
| Text, CSV, JSON, XML, logs, SQL dumps | 70–95% | Massively repetitive; the ideal case for Deflate |
| Word, Excel, PowerPoint (.docx, .xlsx, .pptx) | 0–5% | Already ZIP archives internally — you are zipping a zip |
| 0–10% | Streams inside are already Flate-compressed | |
| JPG, PNG, WebP, HEIC, AVIF, GIF | 0–2% | Already compressed; no repetition left to find |
| MP3, MP4, MOV, MKV | 0–1% | Same reason — the codec did the work |
| BMP, WAV, uncompressed TIFF | 50–90% | Raw pixel and sample data, highly repetitive |
| Executables, DLLs | 40–70% | Code has a lot of structural repetition |
| Already-zipped files | Slightly negative | Adds headers with nothing to gain |
The pattern is simple: compression works by removing redundancy, and a format that already compresses has none left. This is the complete explanation for why zipping a holiday photo folder saves nothing, and why zipping to email photos is pointless unless you need them in one attachment.
What ZIP passwords actually protect
ZIP has two encryption schemes and they are not remotely equivalent:
- ZipCrypto, the original from 1990, is broken. There is a well-known plaintext attack that recovers the key in seconds given a little known content — and if the archive contains any file whose contents are predictable, that condition is met. Treat it as no protection at all. Many tools still default to it for compatibility.
- AES-256, added in 2003, is genuinely strong, and its weakness is your password rather than the cipher. Not every extractor supports it: 7-Zip and WinRAR do, and the built-in extractors in Windows Explorer and macOS Archive Utility historically do not, so your recipient may be unable to open it.
Two things remain true even with AES. File names are not encrypted in standard ZIP — the central directory is readable without the password, so anyone can see that your archive contains redundancy-list-final.xlsx. And a password emailed alongside the archive protects against essentially nothing. For genuinely sensitive material, use a tool designed for the job rather than a ZIP password.
When zipping helps and when it does not
- Worth doing: sending many files as one attachment; preserving a folder structure that would otherwise be flattened; getting past mail filters that block loose
.exeor.jsattachments; bundling a set of converted images for download. - Not worth doing: shrinking a single photo, video or PDF — compress the file properly with a tool that understands its format instead; squeezing under an email limit when the content is already-compressed media, where zipping buys you nothing.
- Actively unhelpful: zipping files the recipient needs to preview. Cloud storage and messaging apps show thumbnails for loose images and nothing for an archive.
On this site, ZIP Archiver bundles files into an archive and ZIP Extractorlists an archive's contents so you can pull out individual files without installing anything — useful on a locked-down work machine. The Bulk Image Converter returns its output as a ZIP for the same reason: one download instead of two hundred.
Things that go wrong
- Non-Latin filenames turn into nonsense. ZIP predates Unicode; older tools wrote names in the local code page. Modern tools set a UTF-8 flag, but an archive made on an old Windows machine can produce unreadable Thai, Chinese or Cyrillic names on extraction.
- The 4 GB limit. Original ZIP used 32-bit fields. Archives beyond 4 GB need the ZIP64 extension, which modern tools handle and some old ones do not.
- macOS adds extra files. Archives made in Finder include a
__MACOSXfolder and.DS_Storeentries, which look like clutter to Windows recipients. Harmless, but worth deleting before sending. - Zip bombs. A small archive can expand to an absurd size. Any tool that extracts untrusted archives should cap the output — ours does.