Production Technology mined from the well logs. In addition, production history and well completion details (e.g., perforation thickness, volume of total injected fluid and total injected proppant) were acquired for the data-driven analysis. Completion reports and well logs were used to find the top and thickness of the Niobrara. Once thickness was determined, average porosity was obtained from neutron and density-porosity log data in the formation interval. With the porosity information, Archie's equation could be used to calculate water saturation considering FIGURE 1A Locations of Wells In Study Area FIGURE 1B Reservoir Delineation/Drainage Area for Each Well FIGURE 2A Fuzzy Pattern Recognition Analysis (Six Months Cumulative Production) FIGURE 2B Fuzzy Pattern Recognition Analysis (Three Years Cumulative Production) 162 THE AMERICAN OIL & GAS REPORTER resistivity values from induction logs. Once all the data were extracted (reservoir, well design parameters and production), a spatio-temporal database representative of the fluid flow in the reservoir was prepared and processed to be in an appropriate format for pattern recognition technology. This database consisted of static (reservoir properties and well completion specifications) and dynamic (production and days of production in a month) data for each well. In order to make static models and delineate the reservoir, an outer boundary was set and the drainage area for each well was estimated using Voronoi diagram techniques (Figure 1B). The reservoir properties then were populated throughout the area of review using geostatistical methods to generate a high-level geocellular model. Geostatistical distribution maps were generated for all the available reservoir properties collected from well logs. Volumetric reserves were calculated based on geocellular properties and well drainage areas. One of the practical implementations of data-driven analysis is fieldwide pattern recognition, which depicts the areas representing high to low production history and remaining reserves. Fuzzy pattern recognition was applied to determine different sections of the reservoir with varying levels of contribution to production during specified time intervals. This FIGURE 2C Locations of Underperforming Wells (Three Years Cumulative Production)