How to build a metric tree

Most organisations track metrics in isolation. Revenue goes up, but nobody knows which team or lever drove the change. Churn increases, and five teams launch five initiatives without understanding the root cause. A metric tree solves this by mapping cause and effect across your entire business, from your highest-level outcome down to the operational levers that teams control every day.

Building a metric tree is not just a data exercise. It is a model of how your business works. It forces clarity about what actually drives outcomes, who owns each lever, and where effort should be focused. Done well, it becomes the shared language that aligns strategy, operations, and individual contribution.

This guide walks you through eight steps to create a metric tree, from choosing your North Star metric to closing the loop with accountable action. Whether you are building your first metric tree or refining an existing one, each step includes the reasoning behind it so you understand the why, not just the how.

Follow these steps in order. Each one builds on the last, and skipping ahead usually means going back later to fill in what you missed.

1. 1

   Start with your North Star metric

   Every metric tree begins with a single number at the top: the one metric that best captures the value your business creates. This is your North Star. It is the outcome everything else feeds into.

   For a SaaS company, that might be Annual Recurring Revenue (ARR). For a marketplace, Gross Merchandise Volume (GMV). For a consumer app, Monthly Active Users (MAU) or Lifetime Value (LTV). The right choice depends on your business model and stage.

   A good North Star metric has three properties: it reflects real value delivered to customers, it is measurable on a regular cadence, and teams across the company can influence it. If only one team can move it, it is probably too narrow. If nobody can move it directly, it is too abstract.

2. 2

   Decompose into first-level drivers

   Once you have your North Star, break it down into its direct components. This is the first level of your metric tree. Ask: what are the two to four factors that mathematically or causally determine this metric?

   For example, if your North Star is Revenue, the first-level decomposition might be: Revenue = Number of Customers x Average Revenue per Customer. Or for a transactional business: Revenue = Transactions x Average Order Value.

   The key is to make the decomposition reflect how the metric actually works in your business. Do not use a textbook formula if it does not match reality. If your pricing model is usage-based, decompose by usage. If it is seat-based, decompose by seats.

3. 3

   Continue decomposing each branch

   Now take each first-level driver and break it down further. The metric tree grows like an actual tree: each branch splits into smaller, more specific branches until you reach the operational levers that teams control.

   Take "Number of Customers" from the previous step. It might decompose into: Customers = New Customers + Returning Customers - Churned Customers. Then "New Customers" decomposes further: New Customers = Organic + Paid Acquisition + Referral + Partnerships.

   Keep going until you reach metrics that a single team or person can directly influence. For most businesses, that means three to five levels of depth. You will know you have gone deep enough when the bottom-level metrics correspond to things teams do every week: run campaigns, ship features, process tickets, close deals.

4. 4

   Define the relationships

   Not all connections in a metric tree are the same. Some are mathematical identities (Revenue = Price x Volume), some are causal relationships (better onboarding leads to higher activation), and some are correlations that need further investigation.

   There are three main types of relationships to consider: Multiplicative (the parent is the product of its children, e.g. Revenue = Users x ARPU), Additive (the parent is the sum of its children, e.g. Total Users = Segment A + Segment B + Segment C), and Influencing (the child affects the parent, but the relationship is not a clean formula, e.g. NPS influences retention).

   Direction also matters. Does the child metric increase or decrease the parent? Churn has a negative relationship with customer count. Response time has a negative relationship with satisfaction. Getting the direction right prevents teams from accidentally optimising in the wrong direction.

5. 5

   Assign ownership

   Every metric in the tree needs a named owner. Not a team, not a department. A person. Ownership is what turns a metric tree from a visualisation into a system of accountability.

   Behavioural science is clear on this: when a specific person is responsible for an outcome, the probability of action increases significantly. Diffusion of responsibility is one of the most well-documented patterns in organisational psychology. A metric without an owner is a metric nobody is watching.

   Use a RACI model or similar framework to clarify the role for each metric: Responsible (the person who takes action when the metric moves), Accountable (the person who is ultimately answerable for the result), Consulted (people with expertise who should be involved in decisions), and Informed (stakeholders who need to know when the metric changes but do not need to act).

6. 6

   Connect to live data

   A metric tree on a whiteboard or in a slide deck is a starting point. But static trees go stale fast. To make the metric tree useful, plug it into your live data sources: your data warehouse, semantic layer, product analytics, CRM, or even manual inputs for metrics that do not live in a system yet.

   The goal is for every node in the tree to show a current value, a trend, and its relationship to the nodes above and below it. When the CEO looks at revenue, they should be able to drill down through the tree to see which specific operational metric is driving the change they are seeing.

   Start with the data you have. Not every metric needs to be automated on day one. Some metrics will be manually updated weekly. That is fine. The value of the metric tree comes from the structure and the relationships, not from having a perfect real-time pipeline on every node.

7. 7

   Validate with data

   This is where most metric trees fall short. You have built a model of how you think your business works. Now test whether it actually does.

   Use correlation analysis to verify the relationships you defined in Step 4. Does onboarding completion actually correlate with 30-day retention? Does increasing paid spend actually drive new customer acquisition, or is organic doing the heavy lifting? Some relationships you assumed were strong may turn out to be weak. Others you did not consider may turn out to be significant.

   This is not about finding perfect causal proof. It is about having evidence that your model reflects reality. When the data disagrees with your assumptions, update the model. A metric tree should evolve as you learn more about how your business actually works.

8. 8

   Close the loop with actions

   The final step is what separates a metric tree from a dashboard. When a metric moves, the owner is notified. They investigate, decide on an action, and log it against the metric. The action is tracked. The impact is measured. The loop closes.

   This is the accountability layer. It is not enough to see that activation rate dropped 5% last week. Someone needs to own the response. What did they do about it? Did it work? If not, what will they try next? The metric tree provides the structure. The action loop provides the behaviour change.

   Over time, this builds an organisational memory. You can look back at any metric and see the history of changes and the actions that were taken. You learn what works. You stop repeating experiments that already failed. The organisation gets smarter, not just more informed.