What a compression rate actually means

A percentage saved is easy to misread. It depends on what you started with, it can be negative, and a big number is not automatically better. Here is how to read the figure the compressor reports.

After compressing, the tool reports four numbers: the original size, the compressed size, the space saved, and a compression rate. The first three are facts about bytes. The fourth is a ratio, and a ratio is only meaningful next to what produced it.

What the rate is

The compression rate here is how much of the original file is gone, as a percentage of the original. A 4 MB file that becomes 1 MB has a compression rate of 75 per cent. That is the reduction, not the remaining fraction - a distinction that causes a lot of confusion when comparing tools, because some report the ratio the other way round.

It can be zero, and it can be negative

If the compressed file comes out the same size, the rate is zero. If it comes out larger - which happens when a file was already compressed hard - the saving is negative, and the tool says so instead of clamping it to zero.

A tool that always reports a saving is telling you about its reporting, not about your file. The honest case is worth recognising: keep the original when the result is larger.

A bigger number is not a better result

The same photograph can be reported as saving 20 per cent or 85 per cent depending only on the quality setting. Neither figure says which one you want. The percentage measures how much was removed, not how much was worth removing.

Two consequences follow. Comparing compression rates between different images tells you almost nothing, because the images differ in detail, dimensions and starting quality. And chasing a high percentage is how people end up with files that look damaged in exchange for a saving they did not need.

What to compare instead

  • The absolute size. If the destination has a limit, the limit is in megabytes, not percentages.
  • The appearance at the display size. Not at 400% zoom.
  • The dimensions. A file that is 3000 pixels wide when 800 will do has a problem that no quality setting solves. What image size do you need covers that.

Where the numbers come from

They are measured, not estimated: the original file's byte length and the produced file's byte length, read after encoding. If the browser silently returns a different format than the one requested, the operation is reported as a failure rather than saved with a misleading size.

The compressor shows all four figures for every file in a batch, and the guides go further into when compressing is the wrong first move.

Related pages