Metric decomposition

Every business metric is the output of a system. Revenue is not a number that appears out of nowhere. It is the product of how many customers you have, how much each one pays, and how long they stay. Decomposition makes that system explicit. It turns a single opaque number into a set of levers you can reason about, measure independently, and act on.

The distinction between decomposition and simply listing related metrics is important. Listing your top ten KPIs on a dashboard tells you what is happening. Decomposing your North Star metric into its constituent parts tells you why it is happening. One is a reporting exercise. The other is a thinking exercise that changes how your organisation makes decisions.

Decomposition also addresses a deep cognitive challenge. Research in behavioural science consistently shows that people struggle to reason about complex systems as wholes. We are far better at reasoning about components. When you decompose a metric, you reduce cognitive load for everyone who needs to understand it. You replace one hard question ("Why did revenue drop?") with several simpler ones ("Did volume fall?", "Did price change?", "Did churn increase?"). Each of those questions is easier to investigate, easier to assign, and easier to act on.

Not all decompositions are the same. The type you choose determines how precisely you can diagnose changes and how confidently you can predict the effect of an intervention. Understanding the differences between these three types is essential before you start breaking down any metric.

In practice, most metric trees use all three types. The top of the tree is often a mathematical decomposition (revenue into its formula components), the middle uses additive decomposition (breaking each component by segment), and the leaves use causal decomposition (linking operational inputs to the segments they influence). The key discipline is knowing which type you are using at each level and being honest about the confidence it gives you.

Decomposing a metric is not a creative brainstorm. It is a structured process that moves from outcome to drivers and validates each step with data. Follow these steps to decompose any metric in your business.

1. 1

   Start with the outcome metric

   Choose the single metric you want to decompose. This should be a metric that matters to the business but is too high-level to act on directly. Revenue, retention rate, and customer lifetime value are common starting points. Be specific about how the metric is defined and measured before you begin.

2. 2

   Ask what has to be true for this metric to go up

   For the metric to increase, which underlying quantities must change? This question forces you to think causally rather than associatively. If revenue needs to go up, either you need more customers, higher prices, or less churn. Write down every plausible driver before filtering.

3. 3

   Write the equation

   Express the relationship between the parent and its children as a formula. Revenue = New Customers x Average Revenue Per Customer + Existing Revenue - Churned Revenue. If you cannot write an equation, you are dealing with a causal decomposition and should note that the relationship is directional rather than deterministic.

4. 4

   Validate with data

   Pull historical data and check whether the children actually explain the parent. If Revenue = Price x Volume, then multiplying your price and volume time series should reconstruct your revenue time series. If it does not, your decomposition is incomplete or your definitions are inconsistent. This step catches errors that pure logic misses.

5. 5

   Check for completeness

   A good decomposition is exhaustive. If you add up all the children of an additive decomposition, you should get the parent exactly. If you multiply the children of a mathematical decomposition, the same should hold. Any residual means you have a missing branch. Find it before moving on.

6. 6

   Assign ownership and decide depth

   Each child metric needs an owner: the person who is closest to the lever and best positioned to investigate when it moves. Stop decomposing when you reach a metric that a single person or team can directly influence. Going deeper than that creates complexity without adding actionability.

The visual above shows a decomposition of revenue into three branches: new customer revenue, expansion revenue, and churned revenue. Each branch is further decomposed into the inputs that drive it. This structure makes it immediately clear where to look when revenue changes. If expansion revenue is flat, the investigation narrows to upsell rate and cross-sell rate. If churn is increasing, you trace it to churn rate and average churned value. The tree eliminates guesswork and replaces it with structure.

While the principles of metric decomposition are universal, the specific patterns differ by business model. Revenue means something different for a subscription business than it does for a marketplace. The table below shows how the same headline metric, revenue, decomposes into fundamentally different structures depending on how value is created and captured.

These patterns are starting points, not prescriptions. Your decomposition should reflect how your specific business creates value, not a generic template. A SaaS company with usage-based pricing will decompose revenue differently from one with seat-based pricing. An e-commerce business with high return rates needs a branch for net revenue after returns. The test of a good decomposition is whether it matches the reality of your business, not whether it matches a textbook.

Notice that each model has a "key driver to watch." This is the metric within the decomposition that tends to have the most predictive power for the overall outcome. It is the branch of the tree where small changes have outsized effects. Identifying this driver is one of the most valuable outcomes of the decomposition exercise.

Metric decomposition is a powerful tool, but it can be misapplied. When done poorly, it creates false confidence, wasted effort, and a model that looks rigorous but leads to the wrong conclusions. These are the most common failure modes.

Decomposition is not the end goal. It is the beginning. A perfectly decomposed metric that sits in a slide deck or a static diagram is no more useful than a dashboard that nobody checks. The value of decomposition is realised only when it changes how people work, what they investigate, and how they decide where to focus.

The first step from decomposition to action is ownership. Every leaf metric in your tree needs a named owner: someone who monitors it, investigates when it moves, and is accountable for its performance. Without ownership, the tree is an intellectual exercise. With it, the tree becomes an operating system for your organisation. When revenue drops, you do not call a meeting of twenty people. You look at the tree, identify which branch moved, and the owner of that branch is already investigating.

The second step is connecting the tree to live data. A decomposition validated against historical data is a good start, but it becomes transformative when it updates in real time. When people can open the tree and see current values, trends, and anomalies at every node, the tree stops being a reference document and becomes the primary tool for understanding what is happening in the business right now.

The third step is linking metrics to the actions and initiatives that are intended to move them. Every branch of the tree should answer not just "what is this metric doing?" but "what are we doing about it?" This closes the loop between understanding and execution. Decomposition gives you the map. Ownership tells you who is navigating. Live data tells you where you are. Action tracking tells you what you are doing to get where you want to go.

There is a behavioural science principle at work here. Psychologists call it "chunking": the practice of breaking complex information into smaller, manageable pieces so that working memory can handle it. Metric decomposition is chunking applied to business strategy. Instead of asking your leadership team to hold an entire business model in their heads, you give them a structure that lets them zoom in on the part that matters to them right now. Instead of asking a team lead to understand every metric in the company, you give them a clear view of the branch they own and how it connects to the whole. The result is not just better decisions. It is faster decisions, made with more confidence, by people closer to the problem.