In a limited number of case studies, sudden increases in total lightning activity, or so called lightning jumps, have been shown to be related to updraft intensification and severe weather production in convection. In preparation for the planned launch of the next generation NOAA Geostationary Operational Environmental Satellite (GOES-R) Geostationary Lightning Mapper (GLM) in 2014, the lightning jump signature is being developed into an algorithm for the short-term prediction of severe weather, including tornadoes. We are developing the initial concepts, tools, and data base for a rigorous lightning jump algorithm test bed. The purpose is to develop a theoretical framework in which expected algorithm performance can be assessed in a rigorous statistical manner. The framework will take advantage of and yet extend the assessment capabilities of individual severe and non-severe case studies developed using Lightning Mapping Array (LMA) total lightning observations from northern Alabama, Washington DC, and central Oklahoma. The approach is to assume a theoretical statistical distribution for lightning flash parameters (e.g., flash rate (FR), d(FR)/dt (DFRDT) or lightning jump) such as a Gaussian or lognormal for both severe and non-severe storms in order to test the outcome of lightning nowcasting algorithms. We use the bootstrap technique, which is a form of Monte Carlo simulation, to pull the flash parameter samples from the assumed population distributions that are defined by a specified mean and standard deviation. The form of the flash parameter distribution, mean, and standard deviation are guided by the database of available LMA case studies. We have conducted trials on the simplest one-dimensional (1-D) and two-dimensional (2-D) lightning jump algorithms for nowcasting severe weather. The simplest 1-D algorithm is based on either a specified FR or DFRDT threshold that separates non-severe and severe storms. A simple 2-D algorithm uses simultaneous and independent thresholds in both FR and DFRDT. Forecast performance is assessed using a 2x2 contingency table and standard skill scores. Preliminary testing of the lightning jump algorithm performance in the statistical framework has been encouraging. The effect of LMA case study sample size on algorithm development and evaluation has been explored. Finally, the framework is being used to objectively refine the lightning jump algorithm to optimize desired skill scores.