The constant miniaturization of IoT sensor nodes requires a continuous reduction in battery sizes, leading to more stringent needs in terms of low-power operation. Over the past decades, an extremely large variety of techniques have been introduced to enable such reductions in power consumption. Many involve some form of offline reconfigurability (OfC), i.e., the ability to configure the node before deployment, or online adaptivity (OnA), i.e., the ability to also reconfigure the node during run time. Yet, the inherent design trade-offs usually lead to ad hoc OnA and OfC, which prevent assessing the varying benefits and costs each approach implies before investing in implementation on a specific node. To solve this issue, in this work, we propose a generic predictive assessment methodology that enables us to evaluate OfC and OnA globally, prior to any design. Practically, the methodology is based on optimization mathematics, to quickly and efficiently evaluate the potential benefits and costs from OnA relative to OfC. This generic methodology can, thus, determine which type of solution will consume the least amount of power, given a specific application scenario, before implementation. We applied the methodology to three adaptive IoT system studies, to demonstrate the ability of the introduced methodology, bring insights into the adaptivity mechanics, and quickly optimize the OfC–OnA adaptivity, even under scenarios with many adaptivity variables.