What is File Compression?

File Compression: Simple Definition

Compression makes a file smaller by removing redundancy: information that can be reconstructed rather than stored. The essential split is between lossless compression, where the original is recovered exactly, and lossy compression, where some data is discarded permanently in exchange for a much smaller result.

Which one is appropriate depends entirely on what the data is. Lossless is mandatory for a spreadsheet or a program, where a single altered byte is a bug. Lossy is the right answer for a photograph or a song, where the eye and ear will not notice what was removed and the size difference is enormous.

A useful rule of thumb: if a human consumes the data directly, lossy is usually acceptable; if a machine does, it almost never is. Nobody notices the detail JPEG removes from a holiday photograph. Everybody notices when a lossy step corrupts a spreadsheet formula or a line of code.

How Compression Works

Compression algorithms exploit two kinds of redundancy. Statistical redundancy means some symbols occur far more often than others, so they can be assigned shorter codes, the principle behind entropy coders such as Huffman coding and arithmetic coding.^[1]Spatial or temporal redundancy means nearby data is often similar, which dictionary methods like LZ77 (used in ZIP, gzip and PNG) capture by replacing repeated sequences with back-references to earlier occurrences.^[1] Most real compressors combine both: a dictionary stage finds repeated patterns, then an entropy stage packs the result as tightly as the data's statistics allow.

Lossless versus Lossy

Compression divides into two families. Lossless methods let the original data be reconstructed bit-for-bit, which is mandatory for text, executables and archives.^[2] Lossy methods discard information judged imperceptible to achieve far smaller files, and the discarded data cannot be recovered.^[3] Lossy compression is therefore used for photographs, music and video, while lossless compression is preferred where exactness matters.^[2] The decision is not about quality alone but about whether any loss is acceptable at all: a contract or a program must survive compression unchanged, whereas a photo can lose detail the eye never notices in exchange for a file a fraction of the size.

The Quality Versus Size Trade-off

For lossy formats, compression is controlled by a quality setting that decides how aggressively information is thrown away. A high quality setting discards little and keeps files larger; a low setting discards more and produces smaller, visibly degraded results. The art of compression is finding the point where the file is as small as possible while the loss stays invisible, which is why a JPEG saved at around 80 percent quality is often indistinguishable from the original yet far smaller. Pushing past that point trades real, noticeable quality for diminishing size gains.

Why Re-compressing Does Not Help

No algorithm can shrink every possible input; this is a consequence of information theory, since random or already-compressed data contains little redundancy to remove.^[1] Re-compressing a JPEG or ZIP file rarely helps and may even enlarge it slightly, because the predictable patterns the compressor relies on have already been eliminated.^[1] Worse, re-compressing a lossy file (re-saving a JPEG, re-encoding an MP4) applies the loss a second time, a problem called generation loss, so each pass degrades quality a little more. The practical rule is to keep a high-quality master and compress once from it, rather than repeatedly compressing an already-compressed file.

Compression in Everyday Files

Almost every file you use is already compressed in some way. Images like JPEG, WebP and PNG, audio like MP3 and AAC, video like H.264 and HEVC, and document formats like DOCX and PDF all build compression directly into the format. This is why zipping a folder of photos or videos saves almost no space: the contents are already near their compressed size, and a ZIP can only bundle them. The biggest compression gains come from content that is not yet compressed, such as raw text, uncompressed images (BMP, TIFF), or databases, where the redundancy compressors feed on is still present.

How File Compression Works

Lossless compression looks for patterns. If a phrase or a run of pixels repeats, the encoder stores it once and refers back to it, so a long stretch of identical data collapses into a short instruction. This is why a text file compresses dramatically and an already-compressed file barely shrinks: the redundancy has been spent.

Lossy compression works differently. It models perception rather than pattern, asking what a person will actually notice. JPEG discards fine detail the eye is poor at seeing; MP3 discards sound masked by louder sound. The data is genuinely gone, which is why saving a JPEG repeatedly degrades it a little more each time.

Compressing an already-compressed file is close to pointless, and occasionally counterproductive. A ZIP of a folder of JPEGs saves almost nothing, because the redundancy was already spent, and the archive can end up marginally larger than its contents once the container overhead is counted.

Examples of File Compression