Raman intensities from reflection (X(R)) and transmission (X(T)) setups are compared by calculations based on random walk and analytical approaches with respect to sample thickness, absorption, and scattering. Experiments incorporating strongly scattering organic polymer layers and powder tablets of pharmaceutical ingredients validate the theoretical findings. For nonabsorbing layers, the Raman reflection and transmission intensities rise steadily with the layer thickness, starting for very thin layers with the ratio X(T)/X(R) = 1 and approaching for thick layers, a lower limit of X(T)/X(R) = 0.5. This result is completely different from the primary irradiation where the ratio of transmittance/reflectance decays hyperbolically with the layer thickness to zero. In absorbing materials, X R saturates at levels that depend strongly on the absorption and scattering coefficients. X T passes through a maximum and decreases then exponentially with increasing layer thickness to zero. From the calculated radial intensity spreads, it follows that quantitative transmission Raman spectroscopy requires diameters of the detected sample areas be about six times larger than the sample thickness. In stratified systems, Raman transmission allows deep probing even of small quantities in buried layers. In double layers, the information is independent from the side of the measurements. In triple layers simulating coated tablets, the information of X T originates mainly from the center of the bulk material whereas X R highlights the irradiated boundary region. However, if the stratified sample is measured in a Raman reflection setup in front of a white diffusely reflecting surface, it is possible to monitor the whole depth of a multiple scattering sample with equal statistical weight. This may be a favorable approach for inline Raman spectroscopy in process analytical technology.