OK, thanks to Tony Valko and FDibbins I have now found enlightenment on floating point errors. As anticipated I have not accidentally discovered an anomaly which no one had found before. Shame! I still have a problem though. I discovered this "error" when using conditional formatting. I wanted a certain figure to turn red if two other figures were not the same and of course a very minute difference means they are not the same. I can solve this by building in a small tolerance to the formula but it seems amazing that this should be necessary. I have been using this simple formula for years and never had a problem before. Is there a simpler way?