Recently models have been developed that allow to quantitatively describe different aspects of perceptual consequences of cochlear damage. This paper discusses, how such models can be applied in digital hearing instruments. Classical multi-band compression systems aim to compensate for the reduced dynamic range (loudness recruitment) of hearing-impaired subjects. These systems basically represent a very crude and simple model of the signal processing as performed by the healthy cochlea. Sounds are processed and amplified in independent channels with constant bandwidth and a limited frequency resolution. Amplification is solely derived from the physical level in each band. On the opposite cochlear filters are asymmetric, strongly overlapping and their shape and bandwidth is level dependent. Next, the physical level of a sound is not a reliable measure of the resulting loudness percept, e.g. loudness also strongly depends on the bandwidth of a sound. In this paper, a model will be presented that correctly predicts loudness perception for narrow and broad-band sounds in individual hearing-impaired subjects.