The Measurement Hierarchy
A revenue number and a button-click count can sit on the same dashboard, yet answer different questions, update on different rhythms, and demand different responses. The hierarchy sorts them out.
▸ Try the interactive toolThe framework
Not all metrics are created equal. A revenue number and a button-click count may both live on the same dashboard, but they answer completely different questions, update on completely different rhythms, and demand completely different responses from a PM. The Measurement Hierarchy sorts metrics into levels so you know which to watch, how often, and what to do when each moves.
The hierarchy runs roughly from business outcomes at the top (revenue, retention — slow-moving, high-stakes), through product metrics (activation, engagement), down to feature and event metrics at the bottom (clicks, page views — fast-moving, granular). The point isn't that lower metrics matter less — it's that each level serves a different purpose. Confusing the levels is how teams celebrate a click-rate while revenue quietly falls, or panic over a high-level number they can't directly move.
Business outcomes — revenue, retention (slow, high-stakes)
Product metrics — activation, engagement (the bridge)
Feature / event metrics — clicks, views (fast, granular, diagnostic)
Each answers a different question at a different rhythm.
Try it yourself
What it prevents
When metrics from different levels are treated alike, teams optimise granular numbers that don't move outcomes, or stare at outcomes they have no direct lever on.
| The shortcut | What it costs | What it gives you instead |
|---|---|---|
| Flat dashboards | All metrics shown equally; no sense of which actually matter. | The hierarchy clarifies which metrics are outcomes vs. diagnostics. |
| Optimising low-level metrics | Celebrating a click-rate while revenue falls. | Knowing the level keeps low metrics in service of outcomes. |
| Staring at un-actionable outcomes | Watching revenue daily with no direct lever to move it. | The hierarchy points you to the product metrics you can act on. |
| Wrong response cadence | Reacting to a slow business metric like a fast event metric. | Each level has its own rhythm and appropriate response. |
Step by step
Classify your metrics into levels
Sort what you track into business outcomes, product metrics, and feature/event metrics. Most dashboards mix all three flat — separating them is the first clarity.
Match cadence to level
Watch event metrics frequently (they move fast and diagnose), product metrics regularly, and business outcomes less often (they move slowly and shouldn't be panicked over daily).
Connect the levels into a line of sight
Understand how feature metrics roll up into product metrics, which roll up into business outcomes. This is what makes a click meaningful — its link to an outcome (and the basis for the metric tree, Tool 05).
Respond appropriately per level
A moving event metric is a diagnostic clue; a moving business outcome is a strategic signal. Don't over-react to granular noise or under-react to a real outcome shift.
Choose the right level for the decision
Strategic decisions reference outcomes; tactical ones reference feature metrics. Pulling the wrong level into a decision is a common source of confusion.
A short illustration
A team proudly reported a rising click-through rate on a feature — a low-level event metric — and treated it as success. Meanwhile, the business outcome that actually mattered, retention, was slowly declining. The granular metric was up; the metric that paid the bills was down, and the flat dashboard made the two look equally important.
The Measurement Hierarchy would have kept the click-rate in its place: a diagnostic detail, meaningful only insofar as it rolled up into an outcome. Once the team connected the levels, they saw the click-rate gain wasn't translating to retention — the feature was being clicked but not creating lasting value. They'd been optimising a number at the wrong level of the hierarchy.
The leveled metric map
The deliverable is your metrics sorted by level, with cadence and line-of-sight — so each is watched and acted on appropriately.
| Level | Example | Rhythm | Role |
|---|---|---|---|
| Business outcome | Revenue, retention | Slow | Strategic signal |
| Product metric | Activation, engagement | Regular | The actionable bridge |
| Feature / event | Clicks, views | Fast | Diagnostic clue |
Why it matters
Every metric belongs to a level, and the level determines its purpose, cadence, and the right response. A dashboard that ignores levels invites the classic error of optimising a granular number while the outcome it's supposed to serve declines.
The most useful discipline the hierarchy enforces is line of sight — connecting the fast, granular metrics at the bottom to the slow, high-stakes outcomes at the top. Feature metrics are only meaningful as diagnostics for the outcomes above them; an event metric that doesn't roll up into anything is just activity. This is why the hierarchy is the foundation for the metric tree (decomposing an outcome into movable inputs) and the leading/lagging distinction (outcomes lag, inputs lead). A PM who keeps the levels straight watches the right things at the right cadence, reacts proportionately, and never mistakes a clicked button for a healthy business.
Common mistakes
Treating all metrics as equally important. Sort them by level.
Celebrating event metrics while outcomes fall. Keep low metrics in service of outcomes.
Reacting to a business outcome daily like it's an event metric. Match cadence to level.
Tracking feature metrics with no link to outcomes. Connect the levels.
When not to use it
- Very simple products. A tiny product may have few enough metrics that formal leveling adds little — but the principle still guards against vanity metrics.
- As rigid taxonomy. Some metrics straddle levels; the goal is clarity of purpose, not perfect classification.
- Ignoring lower levels entirely. Feature metrics matter as diagnostics — the hierarchy organises them, it doesn't dismiss them.
Where this sits in the toolkit
The hierarchy is how you use data well (Tool 01) — knowing which metrics inform which decisions.
Business outcomes lag; product/feature metrics lead (Tool 03).
Connecting the levels into a line of sight is exactly what the metric tree formalises (Tool 05).
Optimising the wrong level is the root of vanity-metric anti-patterns (Tool 27).
Level a dashboard
Take a dashboard you know. Sort its metrics into three levels: business outcomes, product metrics, feature/event metrics. Notice how many sit at the bottom.
Pick one low-level metric and trace its line of sight upward — does it actually roll up into an outcome, or is it just activity?
If a prominent metric doesn't connect to any outcome above it, you've found a candidate vanity metric — and seen why the hierarchy's line of sight matters.
New tools and AI deep-dives, occasionally.
No spam. Unsubscribe anytime.