© 2019 The Royal Society of Chemistry. The aim of this study is to evaluate the performance of a handheld NIR spectrophotometer for the quantitative determination of two APIs in a solid pharmaceutical preparation (tablet). This is a quite challenging problem since both APIs are present in widely different mass proportions. Paracetamol (PCT) represents approximately 81% w/w of the whole composition, while tramadol (TRA) is only about 9% w/w. Under these conditions, the spectrum of the sample is determined by the spectrum of paracetamol and small variations in the concentration of tramadol are barely detectable. Calibration models and prediction results are compared with those obtained using a well-known benchtop spectrophotometer, using HPLC as a reference analytical method. PLS has been used for building the calibration models by cross-validation. The effect of different preprocessing methods (standard normal variate and derivatives) is also compared. As both spectrophotometers cover different spectral ranges, the comparison has been done in the coincidence spectral zone (1100-1676 nm) using the same preprocessing methods. Additionally, the prediction capacity has been optimized selecting different spectral ranges taking into account the spectra of pure APIs. The results show that by selecting a suitable wavelength range, the handheld spectrophotometer produces results comparable to those from the benchtop instrument, with root mean square error of prediction values of 1.22% w/w and 0.58% w/w for PCT and TRA, respectively. It has been applied for the determination of the uniformity of dosage units of two commercial batches.