It is an increasingly common practice in several natural and social sciences to rely on network visualisations both as heuristic tools to get a first overview of relational datasets and as a way to offer an illustration of network analysis findings. Such practice has been around long enough to prove that scholars find it useful to project networks on a space and to observe their visual appearance as a proxy for their topological features. Yet this practice remains largely based on intuition and no investigation has been carried out on to render explicit the foundations and limits of this type of exploration. This paper provides such analysis, by conceptually and mathematically deconstructing the functioning of force-directed layouts and by providing a step-by-step guidance on how to make networks readable and interpret their visual features.