Building energy simulation is a powerful tool for architects and engineers to utilize in both new building design and in evaluating the performance of existing buildings. Considering that the buildings sector is responsible for approximately 40% of total U.S. energy consumption, being able to properly analyze existing buildings is crucial for our goals of sustainability, environmental stewardship, and energy independence. Building energy simulation applications for existing buildings, including measurement and verification or retrofit savings estimation, require that the model be calibrated with respect to actual performance data (e.g., utility bills) in order for the outputs to be relevant and useful. While there has been research that aimed to make model calibration more of a science, due mainly to lack of software implementation, calibration efforts in practice are still typically based on user judgment and trial-and-error. The large number of possible model parameters to tune allows for calibration to be done arbitrarily. Essentially, there are an infinite number of ways to get a satisfactory match between the model and the performance data, especially when only looking at monthly, whole building consumption values. However, random tuning will most likely not result in a model that matches the building's actual behavior. As a consequence, model outputs at the end-use/sub-system level or at smaller time steps will be inaccurate. In addition, model "fudging" reduces the credibility of the model and leads to the all too common "garbage in -- garbage out" perception of energy modeling. This paper first discusses some of the major challenges that energy modelers face during the calibration process including data collection issues, complex factor interactions, parameter uncertainty, compensating errors, the "identifiability" problem, software limitations, and defining when a model is calibrated. Next, a brief discussion on calibration methodologies found in the literature is presented with some focus on calibration in research vs. calibration in practice. Then, a general and pragmatic calibration methodology that has been followed in practice with success is outlined. This procedure includes the following steps: (i) data collection, (ii) updating and simulating the model, (iii) comparing model outputs to performance data, (iv) targeting model parameters to tune based on uncertainty and sensitivity, (v) performing model diagnostics, and (vi) updating the model based on engineering judgment or returning to step (i) for more data collection. Finally, this paper will present the calibration process and results for a 69,350 ft² (6,450 m²) office and chemical/biological laboratory building.