One of the worst experiences for an intermediate programmer is to debug a sophisticated bit of software only to figure out Python is delivering the wrong data type somewhere in the mix.
I’ve never heard of data science-types using strongly typed languages (TypeScript, Java, C++) for their analyses, but it wouldn’t be the dumbest things I’ve ever heard.
Edit 13 Aug 2019:
I’ve since had a think about my last comments regarding strongly typed languages and data science – and I don’t know what I was thinking. Heaps of DS work is done in strongly typed languages! It is perfectly normal to do analyses in C++, Java, (C!), C#, etc. I think when I wrote that I was thinking, for whatever reason, of TypeScript which is quite young. Hell, I’ve even done work in C#!
Yesterday I talked with a hiring manager about leading an analytics team at a very large organisation. The role was dual-hatted: project manager and people manager. I found the project management component odd, so I asked. ..
Yesterday I talked with a hiring manager about leading an analytics team at a very large organisation. The role was dual-hatted: project manager and people manager. I found the project management component odd, so I asked. She explained the projects were almost always the implementation of websites / pages to display results from the team. Project size was typically a few million dollars each.
With that in mind, now imagine your team delivering analytical results – to an organisation of thousands of people – and not having enough organisational support to actually publish the results. You’d likely be spending a lot of time publishing, presenting, re-presenting, refreshing, and marketing the results. And that would be less time for the next analysis.
I was pretty impressed they supported the analytics team in this way! They weren’t the most tech-savvy folks, but they found a solution.
I suspect the organisation had either:
- Learned this lesson the hard way (i.e. analytics without a strong executive mandate doesn’t have much impact) or
- Has a legislative (or possibly political) requirement to deliver.
Either way, it was great to see a large organisation support analytics in a way that was as concrete as multiple $millions in what is effectively a publishing budget.