Understanding the penetration behaviour of drugs into human skin is a prerequisite for the rational development and evaluation of effective dermal drug delivery. The general procedure for the acquisition of quantitative drug penetration profiles in human skin is performed by sequential segmentation and extraction. Unfortunately, this technique is destructive, laborious and lacks spatial resolution. Confocal Raman microscopy bares the potential of a chemically selective, label free and nondestructive analysis. However, the acquisition of quantitative drug depth profiles within skin by Raman microscopy is impeded by imponderable signal attenuation inside the tissue. In this study, we present a chemical semi-solid matrix system simulating the optical properties of human skin. This system serves as a skin surrogate for investigation of Raman signal attenuation under controlled conditions. Caffeine was homogeneously incorporated within the skin surrogate, and Raman intensity depth profiles were acquired. A mathematical algorithm describing the Raman signal attenuation within the surrogate was derived from these profiles. Human skin samples were incubated with caffeine, and Raman intensity depth profiles were similarly acquired. The surrogate algorithm was successfully applied to correct the drug profiles in human skin for signal attenuation. For the first time, a mathematical algorithm was established, which allows correction of Raman signal attenuation in human skin, thus facilitating reliable drug quantification in human skin by confocal Raman spectroscopy.