I am reviewing my usage statistics for subscription databases (with an eye toward dropping/replacing some of them). Does anyone apply a per-search cost to their analysis to determine cost-effectiveness? If so, how did you arrive at a reasonable cost per search?