We analyze the impact of the magnetic anisotropy on the geometric structure and magnetic ordering of small atomic clusters of palladium, iridium, platinum and gold, using Density Functional Theory. Our results highlight the absolute need to include self-consistently the spin orbit interaction in any simulation of the magnetic properties of small atomic clusters, and a complete lack of universality in the magnetic anisotropy of small-sized atomic clusters. PACS numbers: 36.40.Cg, 71,70.Ej, 75.30.Gw Nanostructures of all kinds display a wealth of fascinating geometric, mechanical, electronic, magnetic or optical properties. The exploding field of Nanoscience pretends to understand, handle and tailor these properties for human benefit. Atomic clusters and chains, molecular magnets, and a number of organic molecules like for instance metallocenes indeed show novel magnetic behaviors, that include the enhancement of magnetic moments due to the reduced coordination and symmetry of the geometry [1], and a rich variety of new non-collinear magnetic structures that are absent in bulk materials [2]. Among all these devices, metallic atomic clusters (MACs) [3] stand out since, on the one hand they represent the natural bridge between atomic and materials physics and, in the other, they can be grown, deposited on surfaces or embedded in diverse matrices, and characterized with relatively well-established techniques. Further, the magnetic properties of MACs show promise for a wide spectrum of applications, ranging from medicine to spintronics.Magnetism in small MACs has been extensively studied both experimentally and theoretically along the past decade [4,5]. It is therefore somewhat surprising that, even though spin-orbit effects are also expected to be enhanced in MACs, there are very few published theoretical papers that include the spin-orbit interaction (SOI) in their simulations [6,7,8,9]. Among these, only a handful have actually looked at the explicit effects of the SOI on the magnetism of MAC. Pastor and coworkers have studied the magnetic anisotropy (MA) [10] of clusters of 3d Transition Metal atoms, using a phenomenological tightbinding scheme. To the best of our knowledge, there are no ab initio studies of the impact of the SOI on the magnetism of 4d and 5d MAC. This includes not only explicit calculations of the magnetic anisotropy energy (MAE), but also and more importantly, whether the SOI modifies the ground state and the tower of lowest lying (magnetic) states of a given cluster.We present in this article a series of calculations of selected atomic clusters of 4d-and specially 5d-elements, that show that the SOI is a key ingredient in any ab initio simulation of heavy transition metal MAC, that should not be overlooked. These include Pd n , Ir n , Pt n and Au n , with n=2, 3, 4 and 5, and also Pd 6 and Au 6 , Au 7 . The rationale behind our choice is that 5d MAC are expected to have the largest MAEs; among all 5d elements, gold is monovalent and should display a simpler behavior; comparing the magnetic p...