We evaluate the performance of small error-correcting codes, which we tailor to hardware platforms of very different connectivity and coherence: on a superconducting processor based on transmon qubits and a spintronic quantum register consisting of a nitrogen-vacancy center in diamond. Taking the hardware-specific errors and connectivity into account, we investigate the dependence of the resulting logical error rate on the platform features such as the native gates, native connectivity, gate times, and coherence times. Using a standard error model parameterized for the given hardware, we simulate the performance and benchmark these predictions with experimental results when running the code on the superconducting quantum device. The results indicate that for small codes, the quasi-linear layout of the superconducting device is advantageous. Yet, for codes involving multi-qubit controlled operations, the central-spin connectivity of the color centers enables lower error rates.