Define or audit a Product North Star, turn it into an exact Metric Contract, expose weak assumptions, and map the inputs and guardrails your team should actually monitor. The tool pressure-tests the decision; it does not manufacture a “best” metric from a score.
North Star Metric Decision Audit
No login. Saved in this browser. Structural analytics track actions and counts, not your metric names, strategy prose, formulas, baselines, targets or evidence notes.
Share privacy: the share link contains the current metric brief. Compression is not encryption; anyone with the link can inspect its contents.
Choose whether you are finding a North Star or auditing one you already use. Unknown is a valid answer; the tool will not silently assume subscription, B2C, weekly cadence or higher-is-better.
Start with the valuable customer change and observable value event. Bring the metric your company already uses, create a candidate from scratch, or adapt starter archetypes. Attention, Transaction and Productivity are optional lenses, not answer generators.
Define entity, segment, aggregation, formula where real, measurement window, direction, inclusion/exclusion rules and source. “Weekly active teams” is not complete until “active” and the eligible team population are defined.
Pressure-test customer value, business link, Product influence, measurability, gaming risk and strategic fit. Keep evidence separate from belief, then record the selected metric, closest alternative, biggest risk and revisit trigger.
A team cannot operate “activation”, “weekly active users” or “GMV” safely from the label alone. The contract records the value event, population, segment, aggregation, numerator/denominator only when applicable, explicit formula when known, window, direction, inclusion/exclusion rules, data source, owner and optional baseline/target.
Formula and direction stay explicit
`Orders × AOV` remains multiplication. A duration can be lower-is-better. A bounded quality metric can use a target range. The tool never infers division because two terms exist or assumes higher is always better.
Missing is better than fabricated
Baseline, target, predictive relationship and data source may stay unknown. The Decision Audit turns those gaps into specific follow-up work rather than filling them with defaults.
The same candidate can be strong on customer value and measurability while still carrying an unsupported business-link hypothesis or a gaming risk. Those are different Product decisions. CraftUp keeps the criterion findings and evidence notes separate instead of summing subjective 1–5 scores.
| Criterion | Question | Evidence state |
|---|---|---|
| Customer value | Does movement represent more of the intended customer value? | Supported / Hypothesis / Unknown |
| Business link | Why should it connect to a sustainable outcome? | Supported / Hypothesis / Unknown |
| Product influence | Can Product influence the conditions without gaming the number? | Supported / Hypothesis / Unknown |
| Measurability | Can analysts calculate the same number reliably? | Supported / Hypothesis / Unknown |
| Gaming risk | Could the number improve while customer experience worsens? | Supported / Hypothesis / Unknown |
| Strategic fit | Can it anchor the current strategy over a meaningful period? | Supported / Hypothesis / Unknown |
Inputs answer: “What factors do we believe affect the North Star and can teams act on?” Breadth, depth, frequency and efficiency remain useful prompts, but the tool also accepts quality, reliability, activation, supply, liquidity and custom lenses. Each relationship keeps its rationale and evidence status.
A guardrail answers: “If we optimize the North Star, what harmful behavior or degraded customer outcome must not get worse?” Define the metric, the harm it protects against, direction or threshold where known, owner and cadence.
One North Star is a useful default for one coherent Product/customer value system, not a law for every company forever. Distinct products, lines of business or customer bases can justify separate anchors. If the team is mixing buyer and seller motions or unrelated product areas, narrow the scope instead of forcing a synthetic compromise metric.
A North Star decision touches several adjacent jobs. This finder owns whether a candidate is a defensible high-level value metric. Everything else routes to the owner built for it.
| Job | Owner | Boundary |
|---|---|---|
| Decompose the North Star into drivers | KPI Tree Builder | The finder exports influence hypotheses; the tree owns formulas, units, denominators, and overlap checks. |
| Commit to objectives and key results | OKR Generator | A key result may reference NSM movement, but the finder never writes OKRs for you. |
| Improve new-user activation | Activation metric guide | Activation is first value, not repeated value; keep it a separate contract and a hypothesized input. |
| Diagnose retention or run cohorts | Retention strategy guide | Retention can be the outcome the North Star predicts, but deep retention diagnosis lives there. |
| Test whether a change moves the metric | Experiment Hypothesis Generator + A/B Test Plan Generator | The finder can nominate experiment outcomes and guardrails; design and analysis stay with the experiment owners. |
| Choose market, positioning, or business model | Product Strategy One-Pager | The metric operationalizes strategy; if strategy is unresolved, the NSM decision may legitimately stay unresolved too. |
Stability is the point of a North Star, teams should not re-pick it every quarter. But it is not permanent. Revisit the decision when the strategy shifts, the primary segment changes, the value event itself changes, gaming is observed, the business-link hypothesis weakens, a guardrail deteriorates, or the metric definition changes. Record one trigger in the decision brief so the review happens for a reason, not on a schedule.
The candidate is DAU, MAU, logins or another raw activity count.
Why: Presence is being used as a proxy for value without a success event.
Do instead: Define the observable value event or threshold and count customers who actually reach it.
A financial metric is automatically rejected or automatically accepted.
Why: The metric name is being judged instead of the value exchange.
Do instead: Ask whether the metric represents delivered customer value for this product and keep the lagging business outcome separate when appropriate.
The highest score wins.
Why: Subjective framework dimensions were summed into false precision.
Do instead: Inspect criterion-specific concerns and evidence, then record the human decision rationale.
Every driver must fit breadth/depth/frequency/efficiency.
Why: A useful heuristic was turned into a compulsory ontology.
Do instead: Use those lenses for prompts and keep Product-specific driver categories when they are more truthful.
The team says a metric predicts renewal without data.
Why: A plausible mechanism was written as a fact.
Do instead: Mark it as a hypothesis until cohort, research or other relevant evidence supports the relationship.
Inputs become an arithmetic tree automatically.
Why: Influence was confused with mathematical identity.
Do instead: Export inputs as influence relationships unless a real formula is explicitly known and modeled.
A North Star is the outcome metric a product or product area chooses to anchor around delivered customer value. KPIs are the broader set of outcome, driver, diagnostic and guardrail measures used to operate the business. This tool focuses on choosing and defining the anchor; the KPI Tree Builder owns the wider metric system.
Use one North Star as a default discipline for one coherent product and customer value system. Distinct products, lines of business or customer bases can justify separate North Stars. If your candidate set mixes unrelated value motions, narrow the scope before forcing a single answer.
No metric label is universally good or bad. DAU or MAU is weak when “active” only means presence and can rise without customer value. An activity metric becomes more defensible when the qualifying activity is a precise value event or threshold that reflects what the customer came to achieve.
Revenue is often better kept as a lagging business outcome, but transactional products can have financial-looking value-exchange metrics such as GMV. The question is not whether the label contains money; it is whether the contract represents delivered customer value and the trade-offs are protected with appropriate guardrails.
Start with a small set of influential, complementary drivers that teams can act on and explain. Three to five can be a useful workshop prompt, not a data-structure rule. A coherent system with two drivers can be valid; a system with six can also be valid if the first-level relationships remain decision-useful.
The selected North Star contract, user-defined input hypotheses and guardrails are exported in a KPI Tree v2-compatible graph. Inputs use influence relationships with evidence status and never become additive or multiplicative formulas simply because they exist. The KPI Tree Builder then owns deeper decomposition, formulas, diagnostics and scenarios.
A counter-metric is any measure you watch so that optimizing the North Star cannot quietly make something important worse, quality, trust, cost-to-serve, or the health of the other side of a marketplace. In this tool every counter-metric is recorded as a guardrail: it keeps its own definition, direction and owner, and it sits beside the North Star rather than being averaged into it.
An activation metric measures whether a new customer reaches first value; a North Star measures repeated value across the customer base you intend to serve. Activation can be a leading input or a diagnostic under the North Star, but optimizing activation alone can fill the top of the funnel while repeat value stalls. Keep them as separate contracts with an explicit hypothesized link.
Yes. Strategy shifts, new primary segments, a changed value event, observed gaming, a weakened business link, a deteriorating guardrail, or a redefined metric are all legitimate revisit triggers. Record one trigger when you decide so the metric is stable by default but never frozen by accident. Changing the definition without versioning it breaks trend comparability, so keep the exact contract with every export.
No. Your metric names, definitions, baselines, targets and rationale stay in this browser (localStorage) or travel inside a share link you explicitly copy. Analytics events carry only structural counts, candidates, inputs, guardrails, finding counts, never your metric text. Anyone with a share link can read the brief it contains, so avoid putting confidential targets in shared links.
If the metric contract is usable, take the selected outcome and input hypotheses into the KPI Tree Builder to build the broader metric system. If the audit says the value exchange or strategic scope is unresolved, resolve that first rather than fabricating a metric to finish the worksheet.
The Finder owns the decision and artifact. The educational article owns the deeper North Star framework; the metrics topic routes to broader measurement and experimentation learning.
Last updated: 2026-09-23