Why an archive needs room to describe itself
A compressor usually replaces repeated sequences with shorter references. If a file contains the same byte pattern many times, the program can store the pattern once and point back to it. But random-looking bytes have no reliable repetitions to exploit. A compressor may still test possible substitutions; when none save space, the encoded result can be about as large as the original—or larger.
The format also has to carry instructions for whoever opens the archive. Depending on the format, that can include markers identifying the compression method, the lengths of data blocks, checksums used to detect damage, or a record of the original file’s name. These details take bytes even when the file’s contents do not become smaller.
Why some files grow by a few bytes
Many compression schemes process data in blocks. A block that compresses well can be stored in its compact form; one that does not may be stored as-is, with a flag or header telling the decoder how to handle it. That lets the archive remain usable without pretending random data has a pattern. The flag and block description still occupy space, so the stored version can exceed the input slightly.
There is a basic limit behind this: no lossless method can make every possible file shorter. If every possible input were mapped to a shorter output, there would not be enough distinct short outputs to represent them all uniquely. Compression succeeds on files with exploitable structure, not on randomness itself.
That is why a tiny increase is not necessarily a malfunction. It may simply be the cost of recording, accurately and reversibly, that the data was best left alone.
