A typical analysis might include a step where the downscaled products derived from various GCMs -- or maybe the raw GCM outputs? --are compared to one another, to determine which GCM performs best for the given area of interest.
We have some basic reporting built into the workflow now, but no metrics for evaluating GCM performance/skill.