Population size is a crucial parameter for both ecological research and conservation planning. When individuals are aggregated, estimating the size of a population through sampling raises methodological challenges, as the high variance between sampling units leads to imprecise estimates. Choosing the right sample design depending on the population aggregation level could improve the precision of estimates; however, this is difficult because studies comparing sample designs for aggregated populations have been limited to a few populations and sampling designs, so their results cannot be generalised. To address this gap, we combined simulations of spatial point populations and field counts of three plant species to compare the relative precision of estimates between three sampling methods: simple random sampling (SRS), systematic sampling (SYS) and spatially balanced sampling (SBS). Comparisons were performed on density and aggregation gradients for a range of sample sizes. Our simulations showed that SYS and SBS were always more precise than SRS when individuals were aggregated, reducing sampling variance up to 80% and 60%. The highest precision for estimating population size was always obtained when the average distance between sampling units equalled the diameter of the clusters (i.e. the groups of individuals). The difference in precision was similar for the natural populations, with sampling variance lowered by up to 75% (SYS) and 60% (SBS) compared to SRS. These findings lead us to recommend using SYS or SBS rather than SRS to estimate population size when individuals are spatially aggregated, as these consistently provide more precise estimates. Assessing cluster diameters in the field enables a quick assessment of the potential gain in precision to expect, and thus the best choice of sampling method depending on the trade-off between precision and field constraints.