A near field targets localisation method using parallel factor (PARAFAC) decomposition with a bistatic MIMO system is presented. PARAFAC has the interesting properties of uniqueness in the decomposition of tensor constructed from the received signal and automatic pairing of emitter and receiver directional vectors of a bistatic MIMO radar, which allow the estimation of the parameters of each source individually. Furthermore, the proposed method provides very good estimation performance, because it does not require the approximation of the directional vectors contrary to the most existing techniques.Introduction: The most existing near field sources localisation methods use an approximated model to overcome the nonlinearity of the arguments of the elements of the directional vectors [1-3], which allows to greatly reduce the computational time but sacrifices precision.A few decades ago, tensor decomposition techniques like canonical decomposition and parallel factor analysis (PARAFAC) were developed to solve psychometric and chemometric problems [4,5]. Unlike matrix decomposition, the tensor decomposition is unique [6]. Another advantage of PARAFAC is its automatic pairing of directional vectors [7], which allows to isolate the contribution of each source and estimate its parameters independently from the other sources.For a near field source, its spatial coordinates provide the minimum and necessary information for its localisation. Thus, this Letter focuses on the joint estimation of the abscissa and ordinate of targets using PARAFAC with a bistatic MIMO system. The novelty of this method is that it avoids the approximation of near field receive and transmit directional vectors, which in return provides better precision.