Optimization of the mechanical aspects of a heated conical oxygen sensor for desired performances, such as low heater power, good poison resistance, fast light-off, and broad temperature range, etc. was achieved with computer modeling. CFD analysis was used to model the flow field in and around a sensor in an exhaust pipe to predict the convection coefficients, poisoning, and switching time. Heat transfer analysis coupled with electrical heating was applied to predict temperature and light-off time. Results of the optimization are illustrated, with good agreements between modeling and testing.