Data is mostly spatial so sometime with ArcGIS or QGIS, or R depending on what exactly it's about; R for statistical properties of data sets, scatter plots for variable relationships etc. Or Matlab for people with engineering backgrounds. I don't generally care what tools people use and everybody has their own experiences and background. Never need nosql databases (or RDBMS for that matter) for our type of work.

I'm experimenting with physical notebooks myself since a few weeks actually. No electronic note taking has ever really worked well for me, although I've been getting by for 15 years. I'm not sure I'll ever find a system I'll really like - it always feels that as long as I find a way that forces me to get intimately familiar with data sets (to the point where I'm re-doing or at least re-thinking the ways the data was made to begin with), the insights bubble up by themselves. In other words I've come to the (regrettable) conclusion that methods and tools don't matter that much, it's the elbow grease that does. (of course I'm not claiming that I could analyze 15gb of data spread over 50 tables with Notepad...)