Kiro dreams of data
A story about sushi and data analysis best practices
This post is adapted from a talk I gave at the dbt Coalesce conference, which employs a fictional sushi master named Kiro as a metaphor for data analysis best practices.
The content and title of the talk is inspired by Jiro Dreams of Sushi, a documentary about a real sushi master and his three-star michelin restaurant in Tokyo.
How does one become a master data analyst?
Does it require technical skills? Should you use pie charts or not?
Perhaps the answer lies in a small village by the sea, where an old sushi chef named Kiro makes his daily 5am stroll to the fish market.
Know your ingredients
Most people walk around the fish market without purpose. They don’t question where the fish came from or how it was caught. They might not even know the difference between shrimp and tuna.
Kiro is different. A lifetime of sushi making experience has taught him that to know excellent food is to know your ingredients.
Kiro can tell you the difference between prawns, shrimp, and any other fish at the market. But Kiro’s expertise doesn’t stop at fish taxonomy. He goes out of his way to actively inspect every piece of fish before using it. As he shops for the best cut of tuna, he interrogates each vendor:
Is this Pacific or Atlantic caught?
What is this catch’s toro ratio?
Did this fish live a good life?
Deeply understanding your ingredients is where true mastery begins. As a data analyst, do you actively inspect each table before you start using it?
Data tip #1: Know your source data before you query it.
Actively test implicit assumptions you may have about the data e.g. uniqueness and/or non-nullity on an id column. Make the connection between a row in that table and a specific business process.
Record your recipes
Kiro brings the tuna back to his restaurant. It’s 7 a.m. and the restaurant doesn’t open for another ten hours. Still not enough time. The rice needs to be replenished and the traveling sake merchant is only in town for one more day.
On his way out, Kiro leaves a note for his apprentice.
How is Kiro confident that his apprentice will prepare the fish to his high standard? Because Kiro knows that when you record your recipes to the level of clarity that he has, even a junior chef can replicate his work.
Data tip #2: Document your data cleaning steps.
If you find yourself doing the same cleaning logic query after query, create an intermediate view / table so you can reuse the logic. Better yet, version control that common transform via dbt and dbt docs.
Master egg sushi
After Kiro’s apprentice has finished cleaning the fish, she moves on to crafting the perfect egg sushi. At Kiro’s restaurant, they don’t even let you touch the fish until you’ve mastered this basic dish. They understand that true craftsmanship takes years of hands-on experience and can never be replaced by a 1-day sushi bootcamp.
Egg sushi may seem bland at first glance, yet it’s proved to be the most popular dish at Kiro’s restaurant. Why is that? Because the simplest things done right are oftentimes also the best.
Data tip #3: Favor simple analytical approaches over complex ones.
Master the humble line chart of a clearly-defined metric before investing time in complex machine learning models.
We’ve arrived at the three rules for mastering the craft of making sushi:
Know your ingredients
Record your recipes
Master egg sushi
Yet these rules apply to any craft, data analysis being no exception:
Know your source data
Document your data cleaning steps
Favor simple analytical approaches over complex ones
Remember these rules and perhaps you can be the Kiro of data analysis.