Detection and correct evaluation of low resistivity pay zones using conventional induction logging tools is a major challenge in hydrocarbon exploration. Existing openhole induction logging tools are comprised of transmitter and receiver sensors, with their axes aligned parallel to the borehole axis. If the apparent formation dip is small, the induced currents flow mainly parallel to the bedding planes, thus measuring the horizontal resistivity of the formation. However, many geologic formations exhibit resistivity anisotropy, i.e., the resistivity varies with direction. For example, in thinly laminated sand/shale sequences, where the sand is hydrocarbon bearing, the vertical resistivity measured perpendicular to the bedding is larger than the horizontal resistivity. The low resistivity shales dominate the horizontal resistivity while the vertical resistivity is more sensitive to the more resistive sand layers. Existing induction tools cannot accurately detect and delineate this type of low-resistivity reservoirs and the measured resistivity will be biased towards the low resistivity shales. In exploration wells that are often drilled vertically, or close to vertical, hydrocarbon-bearing sand/shale formation are often overlooked. In collaboration with Shell Technology EP, Baker Atlas developed, built and field-tested a new multi-component induction logging tool to resolve the formation parameters of electrically anisotropic reservoirs. The newly developed induction instrument comprises three mutually orthogonal transmitter-receiver configurations that acquire tensorial magnetic field data. The optimum tool design was based on an extensive resolution study. We analyzed more than 500 different benchmark models comprising 1-D and 2-D structures commonly encountered in hydrocarbon exploration. Part of the study aimed at gaining more insight into the underlying physics governing the complex responses of the new horizontal magnetic field sensors. The sensitivity of the data to formation parameters and the signal-to-noise ratio are used to evaluate our experimental design. We use the singular value decomposition of the sensitivity matrix to map uncertainties in the data into regions of uncertainties in parameter estimates, establishing a link between the uncertainty in the data and the confidence intervals of the interpreted parameter. To realistically quantify the resolution, we developed a comprehensive noise model, which combines thermal noise of the proposed acquisition scheme and systematic noise sources. We also analyzed the sensitivity of the new tool design with respect to environmental noise sources, e.g., borehole rugosity, eccentricity, "yo-yo" effects, tool bending, finite coil lengths, etc. Given the synthetic responses over the wide range of benchmark models and the sensitivities to the formation parameters, we could realistically compare and select an optimum tool configuration to resolve 2-3 m thick anisotropic beds in the presence of 5% data noise. The new tensor induction logging tool was deployed in various field environments and the data that were acquired and interpreted confirmed the predicted capabilities of the new tool.