Compression Ratio

Compression ratio is the original size divided by the compressed size, so 100 MB down to 25 MB is 4, written 4:1. That is the convention in storage and backup. Be careful, because the other convention exists too: some tools report the reciprocal as a percentage, so the same file might be described as "25%", and a few report the space saved, which is 75%. Three numbers, one file, and they are routinely mixed in the same vendor document.

The relationship between ratio and saving is nonlinear in a way that matters for buying decisions. Space saved is $1 - 1/\mathit{CR}$, so going from 2:1 to 4:1 takes you from 50% saved to 75%, a real gain. Going from 10:1 to 20:1 takes you from 90% to 95%, and going from 100:1 to 200:1 buys you half a percent. Past about 10:1 further compression is nearly worthless in storage terms while costing exponentially more CPU, which is why the highest settings in every compressor are a trap.

The floor is set by entropy, not by the algorithm. Data that is already compressed, encrypted or random has an entropy near one bit per bit and will not shrink at all. Try it and most compressors return a file slightly larger than the input, because they still have to write a header, and a compressor that never expanded any input would be a proof that infinite lossless compression exists. Deduplication ratios quoted by backup vendors are a different animal again: they count identical blocks across many machines, which is why 20:1 is plausible for a hundred similar servers and impossible for one.