The merit order stack is used to tackle a wide variety of problems involving electricity dispatch. The simplification it relies on is to neglect dynamic issues such as the cost of starting stations. This leads the merit order stack to give a poor representation of the hourly pattern of prices and under-estimate the optimal level of investment in both peaking and inflexible baseload generators, and thus their run-times by up to 30%. We describe a simple method for incorporating start-up costs using a single equation derived from the load curve and station costs. The technique is demonstrated on the British electricity system in 2010 to test its performance against actual outturn, and in a 2020 scenario with increased wind capacity where it is compared to a dynamic unit-commitment scheduler. Our modification yields a better representation of electricity prices and reduces the errors in capacity investment by a factor of two.