This tool measures the error introduced by the anonymization method for queries that count the number of distinct users.
The tool generates a set of queries across a variety of columns and different column values. The tool also ensures that queries with a variety of different count magnitudes are used for the measure. In other words, small counts, large counts, medium counts, and so on. It measures both absolute error and relative error.
The tool is implemented in the python code gdaUtility.py, found at https://github.com/gda-score/code/tree/master/utility. The same tool also provides a coverage measure.