A statistical model with a strong correlation has been developed to determine average fracture density (FD) on the basis of production variability of 271 individual wells producing exclusively from the Nikanassin and equivalent formations in a large area of more than 15 000 km(2) in the western Canada sedimentary basin (WCSB) in Alberta and British Columbia. Fractional production-variability plots (FPVPs) published by Nelson (2001) have been used successfully in the past in many naturally fractured reservoirs around the world. Up to now, the generation of such graphs has been based on empirical observations from field-production data. The variability of the graphs is interpreted qualitatively as a measure of reservoir heterogeneities. This paper presents a sequential methodology to reproduce empirical fractional variability plots of the Nikanassin tight gas formation using real data, an empirical variability-distribution model (VDM), and dual-porosity numerical simulation. Different simulation approaches and multiple sensitivities were generated from the simplest to the most-detailed cases to understand what causes the curvature of the FPVP in Nikanassin wells. The base simulation case is a homogeneous dual-porosity model, where porosity and permeability are set constant for matrix and fractures. A second case accounts for a heterogeneous dual-porosity model generated through statistical distributions of porosity and permeability. Finally, discrete-fracture-network (DFN) methodology is used to generate multiple fracture models from which stochastic fracture properties are generated for the simulation model. It is concluded that curvature of the variability plot is affected mainly by the occurrence of natural-fracture density. This finding permits estimating FD approximately parallel to the northwest/southeast-trending thrust belt of the Canadian Rocky Mountains in both the west and the east side of the deformation wedge. An unexpected result is that for a significant change in fracture permeability and porosity with constant FD, the fractional production-variability curve is not affected significantly in the case of gas reservoirs. Although the method is applied specifically to the Nikanassin tight gas formation, the theory is developed in detail in such a way that the methodology can be applied in other tight gas reservoirs around the world. Findings from this work are in good agreement with the geological description of the Nikanassin formation and with a production analysis performed in six Nikanassin study areas based on the cumulative number of wells vs. yearly cumulative gas production.