Machine-to-Machine (M2M) paradigm apply to systems composed by numerous devices sharing information and making cooperative decisions with little or no human intervention. The M2M standard defined by the European Telecommunications Standards Institute (ETSI) is the only one providing an end-to-end view of the global M2M architecture. Noticeably, it furnishes a standardised framework for inter-operable M2M services that satisfies most of M2M modelling requirements. However, and even though M2M systems usually operate in highly evolving contexts, this standard does not address the issue of system adaptations. It is furthermore unsuitable for building self-managed systems. This paper introduces a multi-model approach for modelling manageable M2M systems. Said approach consists in a formal graph-based model on top of the ETSI M2M standard, alongside bi-directional updates that ensure layer coherency. Its fitness for enforcing self-management properties is demonstrated by designing high-level reconfiguration rules. Finally, its applicability is illustrated and evaluated using a smart-metering application.