The much harder problem is data management and preparation. Anyone with half a brain and a decent visualization tool can create basic graphs - but that doesn't mean they should, especially if the organization doesn't have good data management processes in place.

Issues like data governance, data prep, and data modelling are the major pain points for me. And honestly, developing a useful BI solution is more about culture change than it is technology. If a company has poor data governance, it doesn't matter how whiz-bang their technology is, they're still not going to get useful insight from their numbers.

There is something potentially harmful, or perhaps that needs addressing, about end-user tools growing in expressive power. A good friend who does statistical genetics work once told me "but I don't want every user running their own regressions and drawing nonsensical conclusions from badly prepared data!"

BTW, that is also the reason I not only set alerts, but review data at a daily level when pulling any rolled-up reports of significance.

We're in the middle of doing this ourselves and would like to hear a random HNers thoughts on the subject.

The other problem is that most executives don't even know what kind of graphs/reports they need. Sure, they can have the classic sales by region and stuff, but the really important things are hidden and require a lot of digging and asking to get to.

When something needs to be regularly produced, then and only then should it be moved to a formal BI solution.

People have a lot of questions and Excel is an approachable tool for small analysis that users can self-serve with.

everybody knows the correct answer is jupyter notebooks, and totes everyone like uses them, that's what it said on my facebook the other day.

NLP interfaces in reporting work for simple things, when you want to start doing more advanced things they become incredibly verbose and inconsistent.

That's where a well designed UI really shines...and will continue to do so.

structured language > purpose-specific UI > natural language (processing)

Old adventure games were really illuminating with respect to limitations of natural language interfaces. Rather than allowing for the expressiveness of natural language, you simply had the feeling of groping around like a blind child at an easter egg hunt for the few valid operations that existed.

I feel like this using GUI tools for data tasks. It's like being forced to communicate by banging rocks together compared to, say, the expressive power of R or Python.

I wonder if natural language interfaces could leapfrog GUIs if we all up and learned something like Lojban...

Really? That's 3 clicks a way with a well-designed UI.

And what if I have those 3 fields in different tables? Even more verbosity...

P.S: My job has nothing to do with designing UIs, just to dispell any doubt :)

That being said, it's a start. SQL as natural language.

How about we just 'say' it ? At one hackathon, I had the idea of making an Alexa skill that would translate simple voice commands like 'graph column A' into matplotlib functions and then show the graphs. Teammates weren't too excited about spreadsheets though.

But I'm assuming lots of things, I can also envision your scenario playing out to great annoyance.

You don't necessarily need to make noise to speak to computers.