From pedantic to pragmatic: Becoming a useful data scientist
Written by Yuzheng Sun
I learned the word “pedantic” from my teammate Craig and instantly loved it.
It’s a spot-on way to classify a certain type of data scientist: They’re painful to work with, they’re counterproductive, but they are dogmatically supreme, making them unchallengeable. The more they know, the extra painful they are.
We were all pedantic data scientists once — and we may still be. And while we certainly don’t want to behave this way, we can’t begin to help ourselves until we understand
- why we are like this
- what’s the better way to behave
Let’s start by answering the latter: Strive towards being a pragmatic data scientist. Pedantic data scientists are blockers, while pragmatic data scientists are seen as teammates.
Ascertaining the value of a data scientist
Throughout my career at different companies, there was always one constant question: How do we evaluate our business value as data scientists? I asked this question, discussed it, and got asked it many times. Nowadays, I actually find humor in it.
Let’s zoom out: Why do businesses exist? Businesses build things — whether material or intangible — that fulfill customer demand. Whoever assists in doing this creates value for the business. It’s that simple.
So why do we have doubts about finding our business value, especially as data scientists in large companies? It’s simple: There are too many roles and layers and concepts that it’s difficult to map your exact contributions towards the underlying goal. We don’t receive tangible feedback on our work from the market or from our customers. Instead, we have to rely on performance reviews to judge our value.
Performance reviews are a whole topic by themselves, which I’ll save for another day. Just remember that they are proxies for your value at best: They are designed to measure and compare, but they are a microscopic representation of the end goal. The end goal has always been to help your business delight its customers.
Be confident in your value
Data scientists are essential because we have the best tools to understand customers and businesses…It is our job, after all.
I’m not only talking about analyses, charts, models, and experiments, but more fundamentally, we have mental models and scientific frameworks. I’ll cover these topics one by one, with industry experts, in the future.
To serve this purpose well requires a pragmatic perspective. A pragmatic perspective means that we start with reality and use the most fitting tools to better understand our existing realities and baselines.
To better illustrate, below is a comparison between pedantic perspective and pragmatic perspective when facing various choices:
It’s hard for pragmatic data scientists to fail, given our training, skills, and the questions we ask. For myself, my career role model is the mythical handyman who simply knows where the problem is by looking and knocking.
I found that good pragmatic data scientists have the same technique because they have the right mental models to continuously sharpen their judgments — AKA intuitions, business acumen, or product sense.
Once we develop a good mental model with good judgments, our careers are almost on autopilot because we can just “feel” what’s wrong.
Why do we choose to be pedantic?
If being pragmatic data scientists is so great, who is stopping us? Spoilers: It’s ourselves, but we need to diagnose ourselves accurately in order to fix it.
I had the opportunity to do a series of data science career sessions with Roberto Medri, one of the best pragmatic data scientists I have encountered. In one of the sessions, he asked the audience to participate in answering these two questions:
- Raise your hand if you know you shouldn’t be too technical in your presentations.
- Raise your hand if you were too technical in your last presentation.
You should’ve seen the crowd. We all raised our hands twice.
We all know that technical presentations lose our audience’s interest and trust, resulting in less impact and lost opportunities, yet we all do it. Why?
Roberto explained the mentality behind this contradictory behavior: We’re insecure.
We are afraid of being proven wrong — or worse, incompetent. So we build up our armor with domain expertise, things we know we can beat everyone on. In our heads, people can’t prove us wrong in debates. But in reality, these obscure analyses are often just useless.
Becoming pragmatic
Being pragmatic is a mindset, but being good at it requires a fair bit of training. Unfortunately, most of us spent many years in school, which actively taught us to be pedantic. There are really only a handful of random blog posts, videos, and interviews that teach us how to truly be pragmatic.
Our channel is dedicated to fixing this gap.
At Statsig, we firmly believe that our value is linearly correlated to the business value our customers can create with data. To that end, we continuously invest dedicated efforts to finding and reporting data science best practices — and we broadcast them for free.
We win when our customers win.
