Quick answer
A North Star metric expresses the recurring customer value a product creates in a way that should lead to durable business results. Input metrics are three to five controllable factors that describe how value is reached, experienced or repeated. Choose the North Star from the product's value exchange, not from a fashionable list. Build a tree of hypothesized inputs, validate them with evidence, add quality and risk guardrails, assign owners and review cadences, and revise the system when strategy or customer behavior changes. The North Star aligns direction; teams improve the inputs.
What a North Star and input metrics are
A North Star metric is a shared measure of the recurring value customers receive from a product. It should sit close enough to the customer outcome to be meaningful and connect plausibly to durable business performance. It is a direction-setting measure, not a slogan or a complete scorecard.
Input metrics describe the factors believed to increase that value. They may represent how many customers reach it, how deeply they experience it, how often it recurs or how efficiently the journey works. Inputs should be concrete enough for a team to change through product or operating decisions.
The combination matters. A North Star without inputs is distant and difficult to act on. Inputs without a North Star encourage local optimization. Together they form a metric tree that connects team learning to a shared outcome.
Choose a North Star from the value exchange
Begin with the job the customer hires the product to do. Identify the observable moment when meaningful progress occurs, then ask whether repeated delivery of that value should support retention, willingness to pay, referrals or another durable result. The metric should represent value, not merely activity.
A good candidate is understandable, measurable with acceptable delay, influenced by several teams and difficult to inflate without creating real value. It normally works at a stable product or business level. A marketplace may need to represent a successful exchange rather than activity on only one side.
No universal metric is inherently a North Star. Weekly active users can be useful for a habitual collaboration tool and misleading for annual tax software. Revenue may be too delayed or reward price extraction, yet it can be a necessary companion for a transactional business. Context decides.
Build the North Star metric tree
Write the North Star at the top and define the eligible population, qualifying event, period and quality threshold. Then decompose it into a small number of inputs. Breadth asks how many eligible customers reach value; depth asks how much value they receive; frequency asks how often it recurs; efficiency asks how reliably it happens.
Use only branches that explain a distinct mechanism. Three to five top-level inputs are usually enough for alignment, with diagnostic metrics beneath them. Name a team owner for each input but preserve shared responsibility where one journey crosses product, marketing, sales and service.
Draw arrows as causal hypotheses. Historical association can help prioritize a branch, but selection effects may explain it. Customers who collaborate more may already be highly motivated. Research, cohort analysis and controlled experiments progressively strengthen or reject the proposed link.
Value
Define the meaningful outcome customers repeatedly seek from the product.
- What progress do customers make?
- What confirms value was received?
North Star
Select a measurable expression of value tied to sustainable business performance.
- Does it represent customer value?
- Can teams influence it responsibly?
Inputs
Map three to five actionable factors that plausibly drive the North Star.
- What expands or deepens value?
- Which team can act?
Guardrails
Protect experience, trust and economics while teams improve inputs.
- What could optimization damage?
- Where might gaming appear?
Validation
Test relationships and revise the tree as the product and strategy evolve.
- Do inputs predict and cause value?
- What evidence would change the model?
Design actionable input metrics
An input should have a clear intervention attached. If onboarding completion falls, a team can inspect steps, eligibility, latency and errors, then test a change. If an input is so broad that no owner knows what to alter, move down one level until the measure supports a decision.
Balance proximity and importance. Button clicks respond quickly but may be remote from value; retained teams matter but change slowly. A useful input sits between them, such as completing a valuable workflow, and has diagnostics that reveal where the workflow failed.
Specify numerator, denominator and window. Raw counts often rise with acquisition while rates reveal experience quality, but rates can improve as volume collapses. Display both where scale and conversion jointly matter. Include data freshness and sample size before teams react to short-term movement.
- Customer value defined
- North Star qualification explicit
- Economic link plausible
- Three to five inputs mapped
- Inputs have owners
- Causal arrows labelled as hypotheses
- Guardrails selected
- Metric contracts versioned
- Review cadences assigned
- Tree revision trigger documented
Add guardrails and companion outcomes
A single metric cannot describe the full health of a system. Guardrails prevent teams from increasing the North Star or an input by harming quality, trust, accessibility, safety or economics. Examples include complaint rates, failure rates, service burden, unwanted contact and contribution per successful outcome.
Guardrails need decision rules, not ceremonial display. Define an acceptable range, the evidence needed to pause a test and who can make that call. Segment them where average safety can conceal harm to a smaller customer group.
Use companion outcomes when the value exchange has more than one necessary side. A marketplace may track successful matches while separately monitoring supplier earnings and buyer satisfaction. Resist blending every concern into one composite whose movement nobody can interpret.
North Star and input metrics example
The design-platform example rejects file creation as the North Star because creating an unused file does not prove collaborative value. A qualified review cycle includes a shared draft, a genuine reviewer response and an accepted decision, creating a more meaningful but still measurable unit.
The inputs locate intervention points across the journey. Activation expands the eligible base, sharing begins the exchange, reviewer response completes the other side, response time affects efficiency and another cycle indicates recurrence. Guardrails discourage empty comments and notification spam.
A hypothetical collaborative design platform wants teams to finish useful feedback cycles, not merely create more files or invite more dormant accounts.
Define customer value as a draft receiving actionable review and reaching an accepted decision within a working cycle.
Measure teams completing at least one qualified review cycle per month, with qualification rules for real participants and an accepted outcome.
Track activated teams, drafts shared for review, invited reviewers who respond, median response time and teams starting another cycle.
Watch low-quality comments, unwanted invitations, notification complaints, failed access and support cost so teams cannot create superficial cycles.
Use interviews and randomized workflow changes to test which inputs improve completed cycles. Review inputs weekly and the North Star and tree monthly.
This illustrative tree encodes hypotheses. The company should not convert every branch into a target until it understands causal effects and possible gaming.
Turn the tree into an operating cadence
Review actionable inputs at the pace teams can change them, often weekly. Review the North Star over a longer window that reduces noise and respects the customer cycle, often monthly or quarterly. Leadership owns the integrity of the shared tree; teams own learning around their inputs.
Meetings should focus on variance, mechanism and decision. Note what changed, what evidence supports the explanation, what will be tested and which owner will act. Avoid reading every tile aloud. The metric system should shorten the path from signal to accountable learning.
Maintain an annotation log for launches, pricing, campaigns, outages and definition changes. A chart with its history attached prevents the organization from repeatedly rediscovering the same explanation and makes later causal analysis more credible.
Validate the metric relationships
First test measurement validity: does the event really represent the intended customer outcome, and can identities and time windows be trusted? Then examine whether inputs lead the North Star consistently across cohorts and segments rather than only in one aggregate period.
Use qualitative research to understand mechanisms and experiments to test interventions. An input can be predictive yet not changeable, or changeable yet harmful when forced. The goal is not to prove a timeless equation but to build a useful, revisable model of the product's value creation.
Watch for Goodhart's law in practice: when a measure becomes a target, people and systems adapt. Audit edge cases, compare quality distributions and ask frontline teams how the metric is being achieved. Reward learning and outcomes, not mechanical input attainment.
Know when to change the North Star
Stability supports alignment, so do not replace a North Star because one quarter is disappointing. Change it when strategy, customer, business model or core product value changes enough that the old measure no longer represents the future being built.
Inputs should evolve more often. A solved onboarding constraint may cease to be the limiting factor, a new platform surface may add a mechanism, or an experiment may disprove a branch. Version the tree and explain what changed so historical comparisons remain intelligible.
Run a periodic review using customer research, economic evidence and team decisions. Ask what behavior the current system rewards, what it ignores and whether it still predicts durable value. Preserve old definitions for analysis even after the operating view changes.
Limitations and common mistakes
The most common mistake is choosing an easy activity metric and naming it a North Star. Messages sent, searches made or sessions opened can increase while the customer outcome worsens. Add a qualification rule that represents successful value rather than motion.
Another mistake is treating the metric tree as proven causality. Correlated inputs may reflect customer intent, seasonality or acquisition mix. Optimization against them can waste resources or manufacture behavior. Keep assumptions visible and use experiments where decisions are material.
Finally, one North Star can erase important business differences. Distinct products or marketplace sides may require separate value metrics under a portfolio outcome. Unity should simplify coordination without forcing unlike customer journeys into one misleading number.
A North Star aligns teams around customer value. Input metrics turn that direction into testable work, while guardrails keep speed from becoming harm.
Frequently asked questions
What is a North Star metric?
It is a shared measure of recurring customer value that should lead to durable business performance and can be influenced responsibly by the organization.
What are input metrics?
They are actionable factors believed to drive the North Star, commonly describing breadth, depth, frequency or efficiency of value delivery.
Can revenue be a North Star metric?
Sometimes, but revenue alone may be delayed or reward extraction rather than customer value. Many products use a customer-value North Star with revenue and contribution as companion outcomes.
How many input metrics should a team use?
A small top-level set, often three to five, is easier to govern. Supporting diagnostics can sit beneath those inputs without competing for equal attention.
How often should the North Star change?
Only when a meaningful strategy, customer, product or business-model change makes the existing measure unrepresentative. Inputs and diagnostics can change more frequently as evidence develops.
Sources and further reading
- Amplitude: About the North Star Framework ↗Publisher framework for customer value, business results and actionable inputs
- Amplitude: Troubleshooting Your North Star ↗Guidance on actionable inputs, breadth, depth, frequency, efficiency and common mistakes
- Amplitude: Changing Your North Star ↗Conditions for revisiting a North Star as strategy and product context evolve
- Google Research: HEART Framework ↗Primary research on mapping goals to signals and metrics for user-centered measurement