I recently had the pleasure of speaking with John David Ariansen of the How to Get an Analytics Job podcast.
I appreciate that John David scoped my background to get a sense of what topics would be good for us to discuss. One of them — course creation — is something I have my opinions on, but I’ve never expressed them in terms of an answer to the basic question: “What makes a good analytics course?“
Here’s my answer on the podcast:
Because I’ve never written down my thoughts about what makes a good course, I figured this is the place to do it: below I elaborate on my thoughts from the show.
A good analytics course…
…Guides you through the data cleaning
Up to 80% of data analysts’ time is spent cleaning data. While it may be excessive to dedicate that much of a course to data cleaning, but this reality should be acknowledged and even celebrated.
It’s so tempting to skip this part of data analysis in a course — it’s hard to simulate data cleaning in a classroom environment, most teachers don’t like teaching it, and most learners don’t like learning it. It’s not the “glamorous” part of data analysis.
However, a data analytics instructor has an obligation to pull back the curtain for their students and show them what really goes on in a data analytics project. Before all the cool analysis happens, there is a lot of unglamorous work.
…Has a context
The world does not need another dispassionate roll-call of every feature and function in an analytics tool. A good analytics course should have some specific learning outcomes and get the student there as linearly as possible.
There are multiple ways to do almost anything in analytics programs.
What’s important is not that the student learn each and every one of them, but that they begin to build the wisdom to know what tool to use, when. This takes the ability to contextualize tools, which means emphasizing some tools and de-emphasizing others, which means the 20-hour course on everything under the sun is not giving you the wisdom to contextualize.
…Has a deliverable
Just as it’s easier to gloss over data cleaning, it’s much easier for an instructor to nibble and peck at various examples and datasets to extract easy demonstration examples.
However, a good course will take the student through the process of creating their own end-to-end deliverable. After all, creativity is the top of Bloom’s taxonomy.
Jumping from topic to topic and dataset to dataset does not frame data cleaning as a necessary part of the data analysis pipeline, nor does it help students contextualize how different tools fit together into one project, nor does it leave students with a sense that they’ve created anything. It’s a cheap shortcut.
What else makes a good one?
Now it’s your turn — what are the traits of a good analytics course? Or, what are examples of an analytics course that “got it right” and you just can’t say why?
Let me know in the comments.
To get the conversation going, I made a video recap of this post. View it below:
You can download the slides from that video below.
Leave a Reply