In this study, diffuse reflectance spectroscopy (DRS) approach was examined for making fertiliser recommendations in the smallholder cocoa farms of Papua New Guinea (PNG). Soil samples were collected from four provinces of PNG. Soil samples from four different depths (0–10, 10–30, 30–60 and 60–90 cm) of 32 profiles in each of these sited were used to create a database of soil chemical and physical properties. Spectral reflectance values at 1 nm interval covering visible to shortwave-infrared (350-2500 nm) were collected for each of these soil samples to develop partial-least-squares regression (PLSR) models. Soil textural fractions, soil organic carbon (SOC) contents and available N were well predicted by the DRS approach with R2 values larger than 0.75. Moderate to poor estimation efficiencies were observed for remaining parameters. Nevertheless, the estimated soil attributes and their corresponding measured soil parameters were used as inputs to a fertilizer recommendation model of Soil Diagnosis to create fertilizer recommendation for a targeted cocoa yield of 1000 kg dry cocoa beans/ha. Resulting fertilizer recommendations were similar for both of these input sources (measured and DRS-estimated) suggesting that the DRS approach may provide an easy way to create fertilizer recommendations.