All businesses are essentially input-output systems. The primary role of data is to understand and illuminate the equations that underlie these systems.
In this post, we will examine how metric trees can aid in expressing these equations. To illustrate this, we'll use a classic transaction-driven model that is applicable to retail, wholesale, or marketplace businesses.
When building a metric tree, you first identify a north star output metric, such as revenue in this example, and break it down into component metrics. As the diagram shows, you break down the metric into multiple levels before reaching terminal nodes that cannot be further broken down.
Two things to note:
- In this first pass, component metrics have clear mathematical relationships.
- Those working in finance will recognize this intuitively. This is how they naturally think – decomposing a P&L into components that mathematically ladder up.
Now, you need to add two important concepts to completely decorate the tree:
1. process metrics - metrics that aren’t precisely mathematically expressed but are related to the key input/output system of metrics. An example of this would be user engagement with customer service.
2. segments - every metric can be segmented or disaggregated via dimensional attributes. Simple examples include the marketing channel for acquiring new signups, or the various devices used by existing customers.
There may be multiple ways to decompose a metric to represent different business processes that ultimately ladder up to the same metric. For example, orders can be broken down into how new and repeat users convert as in the first diagram, or how carts convert, or how visits convert. Each view represents alternate input levers to affect the outputs.
Organizing a business using metric trees (with segments/drivers) provides a powerful foundation for operating the business. Several high-value planning or operational workflows can now be streamlined or even automated. I will present one such workflow in a follow-up post this week.