One tends to use lossless data compression (text compression) on data that can be compressed a lot, for example log files or trace files. For small amounts of compression, the slightly decreased bandwidth or increased storage is not worth the inconvenience.
What is the threshold? Maybe around 5:1, or compression to 20% of the original size. Include many input data which compress that much or more in a compression benchmark.
Incidentally, most lossy compression of media hits that ratio or better.
Another category of useful compression might be compression that is fast enough to be transparent. This will depend on the bandwidth of other parts of a pipeline that transfers and uses data.