Changing direction...a bit

Changing direction...a bit
Photo by Javier Allegue Barros / Unsplash

It's only been two months since I started datafantic. I've been very happy with the response from the community, and I've even gotten the chance to talk with some of you. I've received many comments from my subscribers with excellent feedback. Thank you for that, I've read every email and I've thought about what you have said.

So far on datafantic, I've focused on doing a data project once a week, which often takes the form of a data story. But this has started to wear me out.

First, I got somewhat disappointed in the ephemeral nature of these analyses. I often put 20+ hours into each article, only to turn around and do it all over again the following week. My process usually goes like this:

  • Find a topic, then find some data. Usually, these are connected as I don't want to start until I know there is some data to look at.
  • Spend 4-10 hours collecting and getting the data figured out.
  • The next 2-4 hours are spent doing some basic analysis and quickly visualizing the data.
  • Spending another 2-4 hours cleaning up the visualizations to fit with the datafantic style.
  • Finally, spending 3-4 hours writing it up and ensuring that everything I'm saying makes sense.

What made things worse was that sometimes the projects didn't work. Just like in the real world, I sometimes started to dig into data only to realize there wasn't really anything there.

I like telling stories with data, but what is starting to get at me is that they aren't very useful. They are intellectually stimulating, but that's it. I want something more out of my work. In short, neverending data stories is getting to me.

I still love data, and I believe many problems can be solved by using data and making the right decisions. But what I want is more freedom to do things a bit differently.

Sometimes I want to build a machine-learning model just to see if it works. Sometimes I want to build a quick app to validate a solution to a problem I'm thinking of. Sometimes I want to do an analysis and dig into some data. If these projects fail, I want to share that too.

I've worked hard to use data from the "West" in the stories so far. This is because I think it is more relatable for the internet audience. I'm not sure I will continue doing this moving forward. I'm based in Mongolia (at least for now), and I'm often motivated by what I see in my environment. So moving forward, you will probably see data from other countries (not the US) more often.

So what do the next two months look like? Here are some of the ideas I have come up with:

  • Can you value a house/apartment using pictures? Build a deep learning model to test it out.
  • Would a reverse job board work where I live? Professionals post their resume/CV and companies apply to hire them. I would test it out by building a quick and dirty MVP.
  • Is it possible to detect AI-generated text with deep learning? Probably not, but I want to try.
  • And anything else that strikes my fancy.

These ideas aren't that similar to what I have been doing, and I realize this might not be what you signed up for. This is why I want to let you know now in case you want to, you know, bail.

In a nutshell, what I want to do is something much more like a collection of data science projects. Also, I may not be posting weekly, but I will shoot for at least two projects a month.

I hope this is something you are interested in, but if not, I get it. Thanks for reading.