As an indicator of innovation, patent data is broadly used in the process of academic research, policy-making and enterprise decision. Under this background, this paper focuses on why patent indicators are used to measure innovation, whether patent can represent innovation, and how to optimize the indicators. And we use the method of literature review and apply a framework of " reasons for using patent to measure innovation – possible issues – solutions”. The reasons for using patent to measure innovation can be summarized as the three following aspects. Firstly, patenting and patent are parts of the innovation process, and it transforms into innovation as long as a patent achieves its potential economic effect. Secondly, the standardized documents and the included abundant information, the huge amount and the developing of（online）database of patent, all of these make it advantageous for using patent data. Thirdly, the stability and universality of patent system is also a key foundation for the wide application of patent data. In corresponding to the three above advantages, there are three types of errors. The first one is selection bias. Although patent is an important part of innovation, it is not the only reflection of innovation and not all patent applications/grants are innovations. The second is sample discrepancy. External discrepancy exists in the patent data of different kinds of samples, and patents themselves are different in innovation level and potential value, thus treating patent without differentiating will result in bias. The third is regime divergence. Patent system has two contrary effects on innovation-encouraging and hindering, and divergence also exists among different nations/regions.In order to reduce these errors, we put forward different solutions accordingly. For the issue of selection bias, the solutions include:（i）apply multi-dimensional indicators to measure innovation, thus more aspects of innovation can be reflected; or（ii）construct a patent-included composite indicator to measure innovation. For the issue of sample discrepancy, the solutions include:（i）group the samples according to industry, technology field, etc; or（ii）differentiate and define the value of different patents. For the issue of regime divergence, the solution is to use patent data under single patent system. Also, we put forward three steps in using patent data to measure innovation. Firstly, search and select patent data, and establish patent database according to the research objects and goals. Secondly, choose a suitable form of the patent indicator, such as the selection of patent types, the selection between application and grant data and between stock and flow data. Thirdly, choose one or more methods to reduce errors in the data and sample condition considered.The main contributions of this paper are:（i）we analyze the theoretical foundation, data characteristics and system background as reasons for measuring innovation by patent, and point out the existing errors in the process, thus providing evidence for choosing a rational indicator of innovation;（ii）we summarize measures in reducing certain errors, which provide solutions in making patent a better indicator of innovation, and we suggest that the selection and construction of patent indicator should be based on the research objects and purposes; and（iii）we put forward suggestions in using patent as indicator of innovation, and the selection of innovation indicators and the application of quantitative instruments should also be based on the research questions.