Battery-free intermittent systems provide an alternative for sustainable IoT deployment with long life and almost maintenance-free nodes. Battery-free nodes are usually powered by energy harvested from their environment and use a capacitor as energy storage. The unpredictable dynamics of harvested energy result in an ON/OFF behavior of the battery free IoT node and the program running on the device progresses intermittently. In the intermittent computing paradigm, a power failure is a normal event, but it causes the loss of volatile memory (CPU registers, including the PC and SRAM content), which resets the progress achieved so far. To prevent such losses, checkpoints are performed frequently, saving the volatile memory content into the non-volatile memory. Upon reboot, the latest checkpoint in the non-volatile memory is copied to the volatile memory to resume the program execution where it had stopped. However, determining the right checkpoint frequency and memory allocation is challenging. Too few checkpoints can hinder progress, while too many lead to unnecessary overhead. As non-volatile memories usually consume more energy than volatile memories, memory allocation can also have a huge impact on a program energy consumption. To address these challenges, we propose SCHEMATIC, an algorithm that determines the placement checkpoints in the source code and the allocation of program variables.