There are three main criteria we'd like to test: consistency, accuracy, and usefulness.
We would like identical queries to return identical results (apart from updates to the taxonomic databases and changes in the taxonomy).
How: unit tests.
We would like the results to reflect the best possible match given all available taxonomic information.
How: Ask taxonomists to double-check our work.
We would like the results to be presented in a format which will be useful to the largest number of users.
How: Create a simple front-end so people can play around with this.