Transmission Raman spectroscopy was used to develop a non-destructive quantitative analytical model for the assay of a crystal dispersion-type ointment containing acyclovir as a model drug with a concentration of 3% w/w. The obtained Raman spectra were pre-processed by applying multiplicative scatter correction, standard normal variate, and first or second derivative by the Savitzky-Golay method to optimize the partial least squares (PLS) regression model. The optimized PLS model showed good prediction performance for 85%, 100%, and 115% label claims, with average recovery values of 100.7%, 99.3%, and 99.8%, respectively. Although the material properties and manufacturing method of acyclovir and white petrolatum were expected to be different from those of the calibration set, the mean recovery value of the commercial product was 104.2%. These results indicate that transmission Raman spectroscopy is a useful process analytical technology tool for product development and quality control of a crystal dispersion-type ointment with low drug concentration.