We describe a simple method for certifying that an experimental device prepares a desired quantum state . Our method is applicable to any pure state , and it provides an estimate of the fidelity between and the actual (arbitrary) state in the lab, up to a constant additive error. The method requires measuring only a constant number of Pauli expectation values, selected at random according to an importanceweighting rule. Our method is faster than full tomography by a factor of d, the dimension of the state space, and extends easily and naturally to quantum channels. DOI: 10.1103/PhysRevLett.106.230501 PACS numbers: 03.67.Ac, 03.65.Wj In recent years there has been substantial progress in preparing many-body entangled quantum states in the laboratory [1]. A key step in such experiments is to verify that the state of the system is the desired one. This can be done using quantum state tomography, or techniques such as entanglement witnesses [2]. However, in many cases these solutions are not fully satisfactory. Tomography gives complete information about the state, but it is very resource-intensive, and has difficulty scaling to large systems. Entanglement witnesses can be much easier to implement, but are not a generic solution since known constructions only work for special quantum states.Here we propose a new method, direct fidelity estimation, that is much faster than tomography, is applicable to a large class of quantum states, and requires minimal experimental resources. Let us first describe the setting of the problem. Consider a system of n qubits, with Hilbert space dimension d ¼ 2 n , and let be the desired state, i.e., the state we hope to accurately prepare. We make two basic assumptions. First, we assume that is pure. However, we do not assume any additional structure or symmetry, so our method goes beyond previous work [3,4] to encompass nearly all of the states of interest in experimental quantum information science (e.g., the GreenbergerHorne-Zeilinger (GHZ) and W states, stabilizer states, cluster states, matrix product states, projected entangled pair states, etc.) in a unified framework. Second, we assume that we can measure n-qubit Pauli observables, that is, tensor products of single-qubit Pauli operators; we do not need to perform any other operations. Thus our method is applicable to any system that is capable of single-qubit gates and readout, without needing to rely on 2-qubit gates or entangled measurements.Our method works by measuring a random subset of Pauli observables chosen according to an ''importanceweighting'' rule. Roughly, we select Pauli operators that are most likely to detect deviations from the desired state . We use the resulting measurement statistics to estimate the fidelity Fð ; Þ, where is the actual state in the lab. Surprisingly, although there are 4 n distinct Pauli operators, we only need to sample a constant number of them to estimate Fð ; Þ up to a constant additive error, for arbitrary . That is, for every possible state , with high probability over the choice...