That's a great suggestion. We will eventually post more detail about this. It will make more sense when divulged at the same time we report on the results of some data quality experiments we are currently running. Stay tuned! :)