Quantum systems are subject to random phase errors that can dramatically affect the fidelity of a desired quantum operation or measurement. Quantum error correction techniques have been developed to facilitate quantum information processing, but the resource requirements are large. This motivates a search for alternative strategies to suppress dephasing in quantum systems. Now Michael Biercuk and colleagues experimentally validate the use of a technique known as dynamical decoupling using optimized pulse sequences to suppress qubit error rates. They find novel pulse sequences that suppress errors by orders of magnitude compared to other existing sequences, and the technique should be applicable across a variety of qubit technologies. Quantum systems are subject to random phase errors that can dramatically affect the fidelity of a desired quantum operation or measurement, but existing quantum error correction techniques have large resource requirements, motivating a search for alternative strategies. The authors experimentally validate the use of the dynamical decoupling technique to suppress qubit error rates, using novel optimized pulse sequences that suppress errors by orders of magnitude compared to other existing sequences. Any quantum system, such as those used in quantum information or magnetic resonance, is subject to random phase errors that can dramatically affect the fidelity of a desired quantum operation or measurement1. In the context of quantum information, quantum error correction techniques have been developed to correct these errors, but resource requirements are extraordinary. The realization of a physically tractable quantum information system will therefore be facilitated if qubit (quantum bit) error rates are far below the so-called fault-tolerance error threshold1, predicted to be of the order of 10-3–10-6. The need to realize such low error rates motivates a search for alternative strategies to suppress dephasing in quantum systems2. Here we experimentally demonstrate massive suppression of qubit error rates by the application of optimized dynamical decoupling3,4,5,6,7,8 pulse sequences, using a model quantum system capable of simulating a variety of qubit technologies. We demonstrate an analytically derived pulse sequence9, UDD, and find novel sequences through active, real-time experimental feedback. The latter sequences are tailored to maximize error suppression without the need for a priori knowledge of the ambient noise environment, and are capable of suppressing errors by orders of magnitude compared to other existing sequences (including the benchmark multi-pulse spin echo10,11). Our work includes the extension of a treatment to predict qubit decoherence12,13 under realistic conditions, yielding strong agreement between experimental data and theory for arbitrary pulse sequences incorporating nonidealized control pulses. These results demonstrate the robustness of qubit memory error suppression through dynamical decoupling techniques across a variety of qubit technologies11,14,15,16.