A quarter-scale laboratory prototype of a centrifugal fertilizer spreader was evaluated to determine whether a conical flow-guiding insert could improve the transverse distribution of granular fertilizer. Experiments were conducted at the Department of Mechanical and Agro-Industrial Engineering, Higher School of Engineers of Medjez el Bab, Tunisia. The methodology combined indoor spreading tests with three replicates per treatment, particle-size classification and reduced-order semi-empirical modelling. The prototype was equipped with a 19 cm inclined disc, three vanes and a conical insert with four longitudinal channels and a 12 mm discharge opening. Transverse profiles were measured from -5 to +5 m without the insert at 779 rpm and with the insert at 708, 779 and 816 rpm, using a fertilizer flow rate of 4.2 kg min⁻¹. Additional half-profile measurements were collected for three particle-size classes of 2.00, 1.25 and 1.00 mm. The conical insert reduced the coefficient of variation from 68.2% without insert to 12.0% at 779 rpm with insert, indicating a marked improvement in transverse uniformity. The best insert configuration was observed at 779 rpm, whereas 708 and 816 rpm produced higher variability. Particle-size analysis showed that the 1.25 mm fraction produced the most uniform half-profile, while 1.00 mm particles were concentrated near the centre and 2.00 mm particles contributed more strongly to distant positions. Because an initial physics-based model did not adequately reproduce the measured profiles, the final modelling strategy used parsimonious semi-empirical functions selected by corrected Akaike information criterion and evaluated by leave-one-out cross-validation. This approach provided a transparent framework for comparing configurations and documenting the effects of rotational speed and particle diameter.