Whereas research primarily focuses on understanding under what conditions sudden transitions in the dynamics and functioning of ecological systems may occur, the scale and complexity of ecosystems limit our capacity to achieve this. Indicators of resilience may help circumvent such limitation by signalling the proximity of ecological systems close to an abrupt transition. However, their successful application strongly depends on the ecosystem under question. Therefore, if we aim to use resilience indicators for ecological management in practice, we need to understand where and how they can be reliably monitored. Here, we test the performance of resilience indicators across species in simple modules of competition to help recognize best-indicator species in a community. We show that differences in species sensitivity to disturbances in a community is affected by the dominant eigenvector of the linearized system at equilibrium, We then use simulated time series to compare trends in variance and autocorrelation across species and at community level. We found high heterogeneity in the strength of the indicators across species, while community-based indicators scored better on average than indicators at species level. Looking at species features, we found that collapsing and invading species showed strongest trends, but we observed no relationship between the number of species interaction links and indicators. Lastly, we explored whether it is possible to identify best-indicator species based on their contribution to community variability using eigenvector decomposition methods Our results suggest that successfully identifying a best-indicator species for critical transitions in multispecies communities is not an easy task.