Our study underscore a key issue in conservation genetics as a recent decline inferred by an unstructured model can be mis-interpreted as a consequence of recent anthropic pressures (Ceballos et al., 2015) when it actually results from meta-population structure. This is all the more alarming since the majority of species is likely organised in meta-populations across their range rather than panmictic at a large scale. We therefore stress the necessity for an educated choice of tools to correctly uncover the recent trend of a species and design proper conservation programs. For instance, detecting a recent bottleneck in meta-populations will require summary statistics measuring the linkage disequilibrium (Boitard, RodrÃguez, Jay, Mona, & Austerlitz, 2016; Kerdoncuff, Lambert, & Achaz, 2020) and/or the inferential framework based on the IICR (Chikhi et al., 2018; RodrÃguez et al., 2018) coupled with whole genome data. On a positive note, we showed that the colonization time of the array of demes can be estimated to some extent (and under some combinations of parameters) by unstructured models. We believe that this is particularly important because it has been shown that the simple instantaneous colonization process we used here behaves similarly to a spatial explicit range expansion (Hamilton, Stoneking, & Excoffier, 2005; Mona, 2017), which is certainly a more realistic model but more difficult to investigate. We are aware that the meta-population models here tested are simple and the parameters chosen are specific of the three shark species we focused on. Nevertheless, the time-scale separation of the coalescence process is general, and it allows explaining intuitively any structured models. The four shark species here used as an example has the merit to cover a large spectrum of LHT and consequently a large spectrum of demographic scenarios, going from a highly structured to a panmictic population: this has strong implications on the distribution of coalescence times and therefore on the interpretation of the observed data.