What you're measuring
The tool compresses your text with the browser's native CompressionStream — gzip (deflate + header/checksum) and deflate (zlib-wrapped deflate) — and reports compressed size, bytes saved and ratio. This is a real measurement of your actual payload, not a guess from a generic table.
Where Brotli fits
- Brotli (br) typically beats gzip by 10–20% on HTML/CSS/JS and is the default for HTTPS when the client advertises it (
Accept-Encoding: br). - Browsers expose decompression for brotli but not compression — so an honest static-site tool can't measure it directly; the reference band (gzip × 0.82–0.9) is the practical expectation.
- Level matters: gzip -1 vs -9 differs ~5%; brotli -4 vs -11 differs ~8% with big CPU cost at the top.
- Small payloads (under ~1 KB) often compress worse than nothing — the format headers alone eat the gains. This tool shows you exactly when that happens.
Reading the table
Ratio is compressed ÷ original (lower is better; 0.35 means you saved 65%). The Brotli estimate row applies the typical band to your measured gzip size. For CDN decisions, compare your gzip ratio here, then check what your host serves via curl -H "Accept-Encoding: br,gzip" -I.