In a recent work [A. Aloy et al., Phys. Rev. Lett. 123, 100507 (2019), 10.1103/PhysRevLett.123.100507] we have considered the characterization of entanglement depth, from a device-independent perspective, in a quantum many-body system. We have shown that the inequalities introduced by J. Tura et al. [Science 344, 1256 (2014), 10.1126/science.1247715] can be used to obtain device-independent witnesses of entanglement depth (DIWEDs) and that they enjoy two key properties that allow one to compute their k -producibility bounds more efficiently for larger system sizes, as well as yielding experimentally friendlier device-independent witnesses of entanglement depth: they involve at most two-body correlators and they are permutationally invariant. While the main aim of our previous work was to illustrate the main ideas and applicability of the method, here we outline the details and complement its findings with detailed analysis and further case studies. Specifically, we consider the problem of finding the k -producible bounds of such DIWEDs under different assumptions. Not surprisingly, with the weakest assumptions, we can compute k -producible bounds only for a relatively small number of parties; however, we can still learn interesting features from these solutions that motivate the search on larger systems under the assumption that these features persist. This allows us to tackle the case where the system size eventually reaches the thermodynamic limit.