Comment

Comment by Tom Rampley on October 2, 2012 at 3:53pm

Excel's not meant for data mining. It's a financial tool, principally, and while I use it to scrub small data sets from time to time simply because I'm so used to using it (coming from a finance background), I don't ever see it becoming competitive with R or SAS, etc. It lacks the functionality, and more importantly, the speed of those programs. Excel will continue to act as a good first tool for analysts in B-school to do simple regressions, learn about optimization (through the use of Solver), and get a feel for certain database-like functions, but for folks whose main goal is to become data scientists I don't see it ever being a big part of their day to day workflow. If MS could somehow figure out a way to layer R over Excel, it could be pretty sweet because as I mentioned it is useful for cleaning up and manipulating small data sets, but as it is with VBA it just has too many deficiencies.

Comment by Miles Garnsey on September 21, 2012 at 7:20am

I don't think Excel is going to be a new giant until they replace VBA with something more powerful. As is, support for VSTO (and therefore C#) is sketchy at best.If I want to create a clustered, stacked bar chart I need to create a VBA script; considering VBA is a language from the 90s and is expected to be retired soon, that doesn't seem acceptable to me.

Excel with powerpivot presents a compelling and powerful data access, ad hoc analysis and pre processing proposition, but as an analysis tool it's a non-player - there are no statistical tests, let alone predictive modelling. Even for serious data manipulation, v-lookups are weak compared to real database joins, and far harder to script than something like linq.

That said, Excel is always going to be one of the most important analysis tools due to it's ubiquity, and is a preferred option for small tasks where a real analytics solution is overkill, I guess it does help us avoid over engineering problems at times.