The assessment of banking performance has been an object of many researches in the field of productivity analysis. Due to the complex nature of banking production processes, the required efficiency measurement approach should provide the possibility of multidimensional assessment. It is achievable using methods of frontier data analysis such as DEA. In the scientific paper there is developed a concept of banking efficiency measurement on the basis of Data Envelopment Analysis assuming deposits as an output. Using the current information technology, there are analyzed dynamics of a set of Latvian banks’ efficiency level, sources of inefficiency, optimal volumes of output and inputs for inefficient banks’, assessed the effects of bank merger. On the basis of the alternative DEA Window analysis approach, there is calculated the average efficiency level of Latvian banks’, which is dependent on the time factor influence. In order to determine the most efficient banks (benchmarks), there are calculated levels of technical, scale and overall efficiency. The author recommends to use these indicators practically in strategical and operational planning, decisionmaking process, benchmarking and risk management.