Techniques for searching for best matches in tables of information are provided. A first tree is traversed according to a first value for a first field and as information is encountered at each node, the information is written to a second tree. The second tree is traversed according to a second value for a second field to determine the best match for the first and second values. By utilizing the first and second trees, efficient searching for best matches can be achieved while allowing for dynamic modification of the data.