Product Metric Governance
One north star, an input tree, and rules so numbers stay honest
A north star metric and an input-metric tree keep a product team aligned. Governance — dictionaries, owners, SLA checks — keeps the numbers believable.
What you will learn
- Define a north star metric with the properties it must satisfy
- Decompose a north star into a tree of input and guardrail metrics
- Apply governance: dictionary entries, owners and freshness checks
Pick one metric that best captures sustained customer value (the north star), decompose it into controllable inputs, and govern every metric with a definition, owner and SLA so the tree stays believable.
What is this?
Product metric governance is the system of an agreed north star metric, a tree of input metrics that decompose it, guardrail metrics that protect against gaming it, and the meta-rules — a dictionary entry, an owner, a freshness SLA and a review cadence for each metric. Done well, every team optimizes a different part of the same algebra instead of different dashboards.
Why do you need it?
- Teams that each optimize their own metric produce a company of conflicting 'wins'.
- A north star channels energy toward actual customer value rather than vanity counts.
- Input metrics make the north star tractable — you can't 'increase retention' directly, but you can move features that feed it.
- Guardrails stop teams from wrecking quality or revenue while pushing their number.
Mental model
A ship's compass and engine room
North star plus input metrics
The compass (north star) says where you're going; the engine-room gauges (input metrics) say how the systems that move you are performing. Neither is useful without the other.
How it works
- North star = the metric that best captures sustained customer value and best predicts long-term growth; it should be leading, simple and understandable.
- Decompose: north star (e.g. weekly activated consuming teams) = inputs such as new signups × activation rate × weekly active share × breadth.
- Derive one input metric per team; the sum of team wins must move the north star.
- Add guardrails — support volume, latency, refunds — that must not regress while inputs are optimized.
- Govern with a dictionary (definition, grain, owner, source), freshness checks, and quarterly metric reviews.
Anatomy of a metric tree.
In production
- A collaboration vendor's north star is 'weekly active shared workspaces'; marketing owns signups, product owns activation, growth owns weekly-active share — all toward one number.
- An ad-tech company pairs its revenue north star with a 'fraud signal rate' guardrail so growth can't come from junk inventory.
- A public-company-adjacent team documents each KPI in a dictionary with an owner and a freshness SLA, so exec reviews read one source of truth.
Best practices
- Choose a north star that leads, not lags: it should move before revenue does.
- Keep the input tree to ≤3 levels — beyond that, teams get lost in leaves.
- Write one dictionary entry per metric: definition, data source, grain, owner, thresholds.
- Revisit the north star annually or on major strategy change; changing it quarterly creates metric tourism.
Common mistakes
- Naming a total (DAU) as north star when a share or a value metric (sticky engaged teams) drives the business.
- Composite north star nobody can read ('engagement-weighted retention efficiency').
- Letting teams optimize inputs without guardrails — classic short-term gaming (cut support, spike activation).
- No owner and no SLA, so every disputed number takes a detective to settle.
Debugging tips
- North star frozen while inputs fly? Reconcile the tree algebra — an input may not multiply into it.
- When two teams dispute a number, the dictionary entry wins if it exists; if it doesn't, write it first.
- Check freshness monitor failures before explaining a metric move.
- If a guardrail starts moving with an input metric, investigate the causal link before cheering the input win.
Interview checkpoint
Quick quiz
Which property is most important for a north star metric?
Guardrail metrics exist to slow down teams chasing their input metric.