What I Read #1
σ-driven project management, culture-market fit, and more
Readers, in this newsletter issue, I’m doing something different, sharing things I read and short blurbs about what I think of them. Sharing what you read has several benefits. Even in the age of algorithmic recommendations, I still think there’s an immense value to manual human curation, for things you can’t readily find on the web, and I hope you find these valuable!
In no particular order:
σ-driven project management: when is the optimal time to give up? by Erik Bernhardsson. I can’t agree more with this post when it comes to data science. How do you know when to put more effort into feature engineering, modeling, and sourcing new data to increase the accuracy of your model, or give up and move on?
Why culture eats strategy by Evan Armstrong. There’s no one winning formula, one way to succeed. Culture-market fit > product-market fit?
Human needs are universal; product solutions are unique by Ami Vora. In my data science career, I’ve worked with a fair share of stakeholders (mostly marketing folks 🤫) who blindly take a successful feature in another product and apply it directly to their own. Don’t do this! Even when you think a product feature will be successful, you should ideally test in production before launch.
BeReal: a detailed history of BeReal’s rise from 0 customers till yesterday by Ali Abouelatta. A walk-through of how BeReal overcame the cold start problem with network effects.
Hate to break it to you, but your team is full of under-performers by Molly G. I believe data science manager’s job is supporting data scientists align on “what” with stakeholders (e.g. what’s the KPI and metric to optimize, what are the underlying assumptions and the core problem we’re solving) so they can focus on “how” (e.g. which algorithms to use, how to experiment and iterate).
We don’t have a hundred biases, we have the wrong model by Jason Collins. A very interesting piece about behavioral economics. Sometimes if you have growing assumption-breaking evidence, it’s time to build an entirely new model.
If you forget as fast as you read, this is for you by Niklas Göke. A few practical hacks to better remember what you read and learn. The best way to remember things is to experience them. What I usually do is take out a piece of blank white paper and write what I learned as if I’m teaching it to myself.
Don't just say hello in chat. This is pretty funny but so true! Have you had anyone who just says “hey” in chat and never asks the question?
And if you missed what I wrote:
Select posts from the archive:
—
Stay tuned for my next post, coming soon! Hint: it’s about personalization and recommendations. I’ve had so much fun writing it, and I’m sure you’ll love it. If you haven’t yet, sign up below to get it straight to your inbox, so you don’t miss it 😎
How do you like this new post? If enough readers like this kind of “what I read” style, I might write it more often, on top of my (monthly) original piece, so let me know what you think by leaving a comment below 👇
If you prefer to give an anonymous feedback instead, click on a link below to vote. This is a reader-supported publication, so your feedback is invaluable:
About this newsletter
I’m Seong Hyun, and I’m a data scientist, machine learning engineer, and software engineer. Casual Inference is a newsletter about data science, machine learning, tech startups, management and career, with occasional excursions into other fun topics that come to my mind.
All opinions are my own.