A structurally bizarre aspect of working with data is realizing that intricately connected user flows and business processes are artificially broken up during measurement only to rely on heroic data modeling efforts to piece it all back together.
In daily organizational workflows, it feels like utilizing chopped up pieces of a whole where the pieces don’t fit as easily like lego blocks or as cohesively like a puzzle.
This is the default pattern in organizations. Take a SaaS business for example: the user flow starts with marketing generating leads, which are then funneled into sales, then onboarding, followed by customer success, and eventually churn or renewal.
All of these steps are deeply interrelated, yet the data around each of these steps is often captured via different APIs in separate systems, with differing levels of accuracy, time grains, and dimensions, making it impossible to track the full user journey without doing extensive work.
To address this challenge, data teams execute valiant data modeling efforts. First, they clean up the raw facts to ensure they are accurate and consistent. Then, they recognize the key entities or dimensions and the associated attributes that give context to these measurements. Once the facts and dimensions are in place, they move on to creating meaningful metrics that represent the business’s key performance indicators (KPIs).
At this stage, many organizations end up with a set of reports or dashboards powered by these data models. But even after going through these steps, we still encounter a major challenge: how do we connect these metrics and dimensions into cohesive, unified models that reflects the entire business process?
The state of the art for this is additional painstaking work in spreadsheets.
This is where I believe metric trees come in as they represent the pinnacle of data modeling. They go beyond the basic elements of facts, dimensions, and metrics. Metric trees model the complete flow of a business process, illustrating how different metrics interconnect and influence each other across various stages of the business cycle.
For instance, in the case of a SaaS business, the metric tree would start with the highest-level output metric, such as revenue, and branch out to show how acquisition metrics (e.g., new customer leads), activation metrics (e.g., onboarding success), retention metrics (e.g., churn rates), and expansion metrics (e.g., upsell and cross-sell) all interconnect.
This structure mirrors the actual dynamics of the business and reflects how changes in one area affect other areas, allowing for a more holistic view of operations.
In short, metric trees represent the end state evolution of data modeling from fragmented measurements to a unified, connected view of the business.
