Why data modeling is a super skill
It’s all about meaning.
People say meaning is the key to life.
It’s also a skill that will make you successful in data modeling. How?
It will put you in a rare group of people that have a unique understanding of the business. This creates trust with stakeholders.
Think about it…
Modeling data is all about giving meaning to meaningless bits and bytes. As a data analyst, data engineer or data scientist, your primary objective is to drive business outcomes.
But that objective is impossible without modeling data first.
Data models need to balance the most technical details of data with how those details impact the strategy of the business.
You have the knowledge to adjust a few lines of code, changing what data means or how it’s stored, which change can provide deep insights that drive millions of dollars in revenue for the business.
As a data modeler you can help leadership bring in millions of dollars in revenue by adjusting a few lines of code.
That’s wild.
And also rare.
Most people in tech companies have specialized roles. Engineers can gather requirements and write code in their sleep. Marketers and salespeople know how to grow a business, but start to sweat when they hear the word ‘programming’.
That’s why people who can balance both technology and strategy are so valuable for a company. They’re trusted because they can make a huge impact.
How do you get good at this?
One way is to use what I call “top-down modeling.”
Start by looking at the outcomes of the business strategy. These are what the business wants to achieve at a high level. They will probably be things like more repeat customers, increased revenue, higher margins, etc.
Think about the metrics needed to help achieve these outcomes. There are two types of metrics: input and output metrics. Output metrics measure the outcome directly or indirectly. Input metrics, on the other hand give leadership what they REALLY want: knobs and levers to adjust the desired outcomes.
Figure out how data needs to be modeled to develop those metrics. For example an input metric to customer churn might be decreased activity on your app or website. In order to model this properly you have to a deep dive into the logs, find the necessary events that measure this activity and add it to a data model.
By modeling this activity properly, you create a reliable early detection signal that can be used as an input for a metric or for a predictive model.
The input/output metrics framework was developed by Amazon and described in more detail in the book Working Backwards by two of their top executives.
Google uses a slightly different framework they call the GSM framework (Goals, Signals, Metrics).
I’ll cover both of them in more detail in future emails.
For now, think about data modeling at your job with this top-down pattern in mind. You’ll start to become a better data modeler and earn more trust in your organization.
Ergest