The performance of renewable energy-conversion systems depends strongly on the background climate and ambient meteorological conditions. However, common approaches to simulating such systems use aggregated climatic data as forcing. Given strong variability in the climate, along with possible future long-term changes, it is important to understand the trade-offs between high-resolution forcing data and representative data. Here we simulate a concentrated solar power plant driven by hourly-resolved reanalysis data using both air and water-based cooling systems. These simulations allow us to analyze plant performance under realistic meteorological variability at two locations. In addition, we simulate the plant under average climatic conditions, where all sub-seasonal and interannual variability has been removed. These simulations are cheaper to run, but do not represent extreme values in the forcing. Our analysis shows that variability in the direct normal irradiance (DNI) is a critical factor for simulating the plant realistically. We show that where (and when) DNI variability is low, average forcing provides a realistic picture of power output. Average forcing can also provide a good estimate of how the plant will perform on average. However, in cases where understanding how robustly the plant performs, it is critical to include realistic variability in the input data.