The disinfection basin of the Khouribga wastewater treatment plant (Morocco), whose effluent is reused to wash phosphate ore, is a plain rectangular tank (60 m × 20 m) fed by a DN800 pipe at 650 m3/h, with the reagent sprayed along one side wall at its head. We used a depth-averaged finite-volume model, checked against analytical solutions and on two grids, to test whether simple internal structures could make this basin work better. The existing basin was compared with layouts of one, two and three alternating dikes (5, 10 and 15 m long) and with a grid of eleven cubic concrete blocks of 3 m side, over three hydraulic retention times. Without internal structures, the inlet jet runs straight to the outlet and the baffling factor t10/T is only 0.10. One 5 m dike placed near the inlet, on the side of the spray zone, sets up an eddy that sweeps the dosing area: the basin-averaged reagent concentration rose by 157% and t10/T reached 0.31. More dikes of the same length added nothing, and longer dikes made both indicators worse. The block grid gave by far the best hydraulics (t10/T = 0.46, Morrill index 3.7 instead of 25.9), yet it did little for the spread of a reagent sprayed upstream of it. Where the reagent enters the basin therefore matters as much as the structures placed in it, and a block grid combined with dosing into the inflow would provide the longest guaranteed contact time.