Growth Engineering

CHAPTER 4Data Modeling

Data modeling is the act of structuring and organizing your data so that it’s consistent, trustworthy, analyzable, and useful for answering questions. At the core of growth engineering is data, and modeling it in a way that reflects reality is essential to driving growth.

Badly modeled data leads to bad results. If the structure of your data doesn’t align with the questions you’re trying to answer, you’ll either get misleading insights or no insights at all. Growth decisions made on flawed foundations often backfire.

Take a simple example: You want to analyze user drop-off in a funnel, but your user identifiers are duplicated across rows. The model never enforced uniqueness. As a result, your funnel numbers are inflated, and your drop-off rates are meaningless.

Or imagine two tables: one containing user click data, the other storing experiment assignments. Both tables use user_id as the primary key—but in one table, users are being assigned new IDs altogether due to a lack of proper identity management. Now you can’t tell which experiment group a user belongs to.

Good data modeling fixes this. When your data models reflect the real-world use cases relevant to them, they become powerful tools in the hands of end users of the data. Product managers can report results better and make better decisions, engineers can effectively debug experiments, and data scientists can trust the analysis of the experiments.

But what makes a good data model?

First, it captures ...