Yeah, I know, not exactly a revelation. We all know it's numbers tend to be off, but at one point I thought maybe they were consistently off, and could be used to predict actual numbers. Out of curiosity I tracked one of my books for eight months and made a chart. The blue are is NovelRank's numbers, the pink is the actual sales.
The two are unevenly off. The issue seems to be that NovelRank can capture only one sale an hour, regardless if there were one or a hundred. The only time NovelRank's numbers come close to accurate if the book's sales are low.