Data often contains repetition and predictability that take up space without adding information. Compression finds those patterns and encodes the data more compactly, so it takes less room to store and less time to send, at the cost of some computation to pack and unpack it.
Some compression is lossless, restoring the original exactly, and suits text and programs; some is lossy, discarding detail a person is unlikely to notice, and suits images, audio and video. Choosing between them is a balance of fidelity against size.