What needs to be finished first
A written product vision or mission statement is available to benchmark value relevance and long-term scope of the metric.
A funnel or AARRR logic is familiar, so the North Star Metric can be embedded in a measurable input tree.
What needs to be ready before start
Whiteboard or Miro board with three columns (candidates, evaluation, input tree); data snapshot from the last 6-12 months; template for metric card (name, definition, formula, source, update frequency).
One owner (Product Lead, GM, or CEO in early stage); two to five people from Product, Data, Marketing, and Customer; one data analyst who can validate values immediately.
Current top metrics; market examples; definition of Customer Value in the product; known vanity metrics to remove; list of active teams and initiatives.
3-4 h workshop, plus 2 weeks validation
Create three columns. Make criteria visible: reflects customer value, correlates with long-term revenue, is measurable in current data pipeline, and is team-influencable, and cannot be gamed by a single action.
The one question this method answers
Which one metric reflects the value the product creates for users and correlates with sustainable growth?
Marker: Phase
| Step | Duration | Action | Hint |
|---|---|---|---|
1Phase 1: Define customer value | 30 min | Define in one sentence which core value the product delivers, referencing Jobs-to-be-Done or an Aha moment. Quote examples from user interviews. | If the sentence contains features instead of outcomes, rewrite it. Customer Value is what the user achieves, not what the product does. |
2Phase 2: Collect candidates | 45 min | Generate 5-10 metric candidates. For each candidate: definition, formula, data source, update rhythm. Examples: active users per week with minimum action, completed orders per month, messages sent between participants. | Candidates must represent value actions, not just activity. Login count is vanity, completed core workflow is value. |
3Phase 3: Evaluate against criteria | 45 min | Score each candidate against five criteria (value, growth correlation, measurability, influenceability, abuse resistance) on a 1-5 scale. Data analyst checks historical correlation with revenue or retention. | If a candidate scores 1 or 2 in any criterion, exclude it. A North Star should score at least 3 across all criteria. |
4Phase 4: Build input tree | 45 min | Set the winner as the root. Add 3-5 drivers as second level, each with formula relationship. For each driver, 2-3 sub-drivers that teams can work on. | If no team owns a sub-driver, the metric is not currently controllable in the organization. Either adjust team setup or choose a different metric. |
5Phase 5: Validation and rollout | 30 min workshop, 2 week data check | Specify dashboard mock-up and owners per driver. Run two-week shadow mode: track metric daily, test against business events. Then make Go/No-Go decision. | Shadow mode reveals data defects and wrong correlations. Without validation the metric is quietly retired after three months. |
What comes out at the end
Metric profile with name, one-sentence definition, formula, data source, update rhythm, owner, baseline, target corridor, and visualized input tree with mapped teams.
Profile with date and version. Archive previous version when definition or formula changes and document reason. Add end-of-quarter review entry to document.
Compact briefing for strategic context, target state, priorities, and non-goals.
Why is this strategic clarification needed now?
What should be different in 6-12 months?
Which decisions should follow from this?
Concrete filled scenario, fictional example
Concrete filled scenario, fictional example
Customer Value: Solo freelancers find a suitable workplace within 5 minutes for the same day.
North Star Metric: Confirmed bookings with check-in per week (BCW).
Formula: COUNT(bookings) WHERE status='checked_in' AND week=current_week
Baseline (KW 18, 2026): 1,240 BCW. Target corridor by 31.12.: 2,500 BCW.
Source: Production DB, hourly refresh, dashboard in Looker.
Owner: Lisa Hartmann (Head of Product).
Historical correlation BCW vs monthly revenue: r = 0.84 (n=18 months). No manipulation possible without actual check-in.
Recognize symptoms and steer against them
An activity number (logins, app opens) is selected without clear customer value.
Check every candidate metric against Customer Value. If user benefit does not improve when the number rises, it is vanity.
Different teams name different primary metrics and nobody can decide.
Exactly one North Star. Other metrics become drivers in the input tree or counter-metrics. If the business truly has two models, create two products, not two North Stars.
A team optimizes the number through actions that do not help users (email spam, dark patterns).
Harden definition: minimum engagement, quality gate, counter-metric (for example complaints, cancellation rate). Add quality counter to make manipulation visible.
Definition is clear but value cannot be reproduced reliably; values vary by source.
Before rollout, run a data-engineering spike: compute the metric in one canonical source and document with versioning in a dashboard. No canonical pipeline means no rollout.
Input tree exists, but no team feels responsible for any driver.
Name each driver owner explicitly. If no owner can be found, redesign team setup or split driver differently.
Done signals checkable in under a minute
Go to the profile for purpose, similar methods, and sources or continue to the next method in the catalog.