In this work, a real-time and in-situ analytical tool based on near infrared spectroscopy is proposed to predict two of the most relevant coffee parameters during the roasting process, sucrose and colour. The methodology was developed taking in consideration different coffee varieties (Arabica and Robusta), coffee origins (Brazil, East-Timor, India and Uganda) and roasting process procedures (slow and fast). All near infrared spectroscopy-based calibrations were developed resorting to partial least squares regression. The results proved the suitability of this methodology as demonstrated by range-error-ratio and coefficient of determination higher than 10 and 0.85 respectively, for all modelled parameters. The relationship between sucrose and colour development during the roasting process is further discussed, in light of designing in real-time coffee products with similar visual appearance and distinct organoleptic profile.