An approach combining laboratory experiments and numerical modelling was used to investigate the behaviour of a floc population during an idealized tidal cycle. The experiment was conducted on suspended sediments at a concentration of 93 mg l−1 collected in the field. It was based on a jar test device to reproduce tidal-induced turbulence and coupled with a CCD camera system and image post-processing software to monitor floc size distribution. At the same time, a 0D size-class based aggregation/fragmentation model (FLOCMOD) was developed to simulate changes in the floc population over the tidal cycle. Experimental results revealed strong variability of the behaviour of microfloc and macrofloc populations with respect to the varying shear rates observed in situ. In particular, the major dependency of floc sizes on the Kolmogorov microscale was confirmed. Time-scale differences were also observed for aggregation and fragmentation processes which led to asymmetrical floc behaviour despite symmetrical tidal forcing. Model results, i.e. average diameter, maximum diameter and floc size distribution, were in good agreement with experimental data with an RMS error between observed and computed average diameters of below 25 μm over the tidal cycle. FLOCMOD was optimized in terms of the time step, number of size classes and size range: only seven classes ranging from 50 to 643 μm associated with a dynamically-adaptable time step were needed to correctly reproduce experimental results, characterized by an RMS error of less than 5 μm with respect to the reference case (100 classes from 4 to 1500 μm). Sensitivity analyses were performed on major parameters or processes: initial floc size distribution, primary particle size, fractal dimension and fragmentation function (binary, ternary, erosion or collision-induced fragmentation). Results showed that initial floc size distribution played a role only during the first aggregation stage. Low variability of the fractal dimension did not significantly modify model results, while larger differences were observed when the primary particle size was changed, especially towards the largest sizes (10 μm). Nevertheless, these two structural parameters had a strong impact on the calculated mean settling velocity with differences of 0.2 mm s−1 compared with the reference case. Different fragmentation functions were shown to significantly modify model results, except for collision-induced shear stress. In particular, combining floc erosion with binary breakup in the shear fragmentation term enabled us to reproduce bimodal distributions, patterns that are typically observed in situ.