The second generation of ground-based gravitational-wave detectors will begin taking data in September 2015. Sensitive and computationally-efficient data analysis methods will be required to maximize what we learn from their observations. We describe improvements made to the offline analysis pipeline searching for gravitational waves from stellar-mass compact binary coalescences, and assess how these improvements affect search sensitivity. Starting with the two-stage ihope pipeline used in S5, S6 and VSR1-3 and using two weeks of S6/VSR3 data as test periods, we first demonstrate a pipeline with a simpler workflow. This single-stage pipeline performs matched filtering and coincidence testing only once. This simplification allows us to reach much lower false-alarm rates for loud candidate events. We then describe an optimized chi-squared test which minimizes computational cost. Next, we compare methods of generating template banks, demonstrating that a fixed bank may be used for extended stretches of time. Fixing the bank reduces the cost and complexity, compared to the previous method of regenerating a template bank every 2048 s of analyzed data. Creating a fixed bank shared by all detectors also allows us to apply a more stringent coincidence test, whose performance we quantify. With these improvements, we find a 10% increase in sensitive volume with a negligible change in computational cost.