I have already mentioned that there are some specific differences between the analytics andsoftware development workflows. Nevertheless a collaborative tool, enabling storing, prototyping and deploying project is a must. Git and Github have been there for the software development folks, but while it is powerful, it is not tailored for analytics. Sense though have recently launched seems to realy aim to fill this gap.
What is Sense?
Sense is basically a data science workbench provinding you access to computing through most open/open source analytical languages, such as R, Python or Julia.
Why is it like Github?
Like Github, Sense provides you with the basic features needed for developing products where the basis is code. So you will have collaboration features, which will allow you to work in teams or to sort of “crowd source” your project.
Issue tracking and management is also a part of sense so these features will help you identify and track bugs plus enables you to keep improving your product.
Last but not least version tracking is part of Sense too. It is not that important or refined feature as in Github, but definitely does the job. I have never seen an analytics project without multiple iterations, so I would consider this feature as a must have.
There is nothing special here though. As I said, these options are there in Github and in many cases in a more refined way than it is in Sense, but what sets this platform apart are the functions I will mention below.
The fit for analytics
I love the fact that there are functions which address some of my ongoing pain points with analytics. It’s need for strong computing power, automatic data refreshment, need for reproducability and deployment of models made it difficult to work on a single platform before. Or at least I did not know about one until now 🙂
Sense allows you to deploy your modells within its own space, plus allows you to pick the computing power best fit for your project. These are things I have struggled with so far, so Sense sorts it out for me in an easy way, which I really think I can utilize.
Once I’m done with creating an analysis I can also schedule it and/or place it into a dashboard. The onboarding process of Sense gives you a pretty cool example of this through the public Wikipedia API.
It is also important to note that Sense is built to support the most popular open source data analysis languages such as R, Python and Julia plus their markdown versions (iPython Notebooks and R Markdown). While Github can handle most of it, its clearly not optimized for these languages.
As iPython and R Markdown is supported, plus you can share your entire project along with your data sources reproducability is in the DNA of Sense. I consider this as a minimum requirement for any analysis platform and it is done very right here.
I’m sure there are other features I did not discover yet, so if you have anything, let me know in the comments!
Sense is free for personal use, so there is nothing to lose trying it. I will do so myself, as I’m learning R and this seems to be a good platform to utilize. If you want something to really easily create and deploy data science projects requiring collaboration, then this could be your Github for analytics.