The ratio between mannitol and sucrose in a freeze-dried formulation has a major impact on the processing and the stability of a lyophilised product. The ratio of these common excipients influences a critical quality attribute of the system, namely the overall amount of water, due to the different nature of their water-solid interactions. For this experiment samples containing various ratios of mannitol and sucrose and several other additives were freeze-dried, stored under different conditions and measured by NIR. Different spectral pre-processing methods and wavelength selections were tested. Multivariate analysis was applied to correlate the Karl Fischer titration to the NIR spectra. It was found that standard normal variate (SNV) transformation of the wavenumber range 4200-7400 cm(-1) yielded prediction errors close to the accepted measurement error of the Karl Fischer titration, when measuring samples of up to 5.5% (w/w) water. It was further found that there was a slight tendency for samples containing inorganic salts or histidine to be underestimated in the NIR measurements. However, no influence was found to be caused by the varying mannitol-sucrose ratio. By reducing the sample set to those samples containing up to around 2% of water, the error was found to be below the uncertainty originating from the reference method. Due to this it can no longer be determined whether the deviation originates from the NIR method or the reference method. It can therefore be concluded the NIR is a suitable tool for quantification of the water content in lyophilised samples with varying mannitol-sucrose ratios.