North Star Metric
beta · Family strategy-docs · Phase undefined · Sizes lean, full · ~3,850 tokens
The document that names the one metric showing customers get the value the strategy promises, says why that metric, and names the input metrics a team can move. Not the vision’s north star statement; the formula and targets belong to kpi-dashboard.
An operator card. Read it before you fill the template and again before you agree the definition. The
reasoning behind every rule here lives in north-star-metrics_companion.md.
Before you start: is this the right document at all?
Section titled “Before you start: is this the right document at all?”A North Star Metric names the single measure of customer value that shows a product’s strategy is working, and the small set of inputs a team can actually move to drive it. If your real question is something else, write something else.
| If the real question is | Write this instead |
|---|---|
| Where are we trying to get to, over years, and why is it worth getting to? | product-vision |
| Which problems will we solve to get there, and which will we not? | product-strategy |
| In what order will we work on those problems? | product-roadmap |
| What measurable change do we expect this period, and did we get it? | okrs |
| What is the formula, the data source, the refresh schedule, or the target behind this number? | kpi-dashboard |
| What does the whole hierarchy of metrics beneath this one look like? | the catalog’s own metric-tree document type, which the companion notes is not yet built in this library |
Write nothing at all if you cannot yet name a product strategy for this product. A North Star measures whether a strategy is moving the product toward its vision; with no strategy stated, there is nothing for the number to measure yet, and the document would be guessing at both ends.
Picking a variant
Section titled “Picking a variant”lean carries the metric’s name and definition, the case for it, its inputs, what it serves, and the trigger that would bring you back to it. That is enough for a first definition, or for a team that is unlikely to be asked what was rejected.
full adds four sections: Candidates Considered, Guardrails, What It Does Not Tell Us, and Definition History. Reach for full once the North Star has already faced, or will soon face, real scrutiny: named alternatives were seriously compared, a stakeholder will ask “why not X”, or the definition has already changed once and the change needs a record.
Guardrails is worth a separate word. It is the one full-only section whose evidence does not come from Amplitude, the publisher this document type is admitted on. Amplitude’s own two documents never use “guardrail” for a measure tracked beside the North Star; the section instead rests on three other sources. Choosing lean and leaving Guardrails out is not skipping an Amplitude step.
The one test that outranks the rest
Section titled “The one test that outranks the rest”Could your team move this number directly? John Cutler, co-author of Amplitude’s own playbook, states the test plainly: “If you can move your North Star directly, it’s probably not a good North Star.” A number your team can push on its own is a lever, not a North Star; the North Star is what the levers are supposed to move.
One boundary check sits beside it, about discipline rather than form: does anything here belong to
kpi-dashboard instead? A formula, a data source, a refresh schedule, a target, a threshold. If you catch yourself writing
one of those into this document, stop and open kpi-dashboard.
The rubric
Section titled “The rubric”Score each 0, 1 or 2. Below 15 out of 22, the document has named a metric without doing the one thing a North Star is for: letting someone else ask, of a real piece of work, “does this move the north star”, and get an answer instead of a shrug. The lean variant scores against fewer rows; see the scope table below.
| # | Criterion | 0 | 1 | 2 |
|---|---|---|---|---|
| 1 | Counts a real behavior, not a lever | No behavior is named, the sentence names an aspiration rather than something that happens, or one team could push the number up directly without any input moving first | A behavior is named, but the actor, the window, or both are missing, or a calculation or data source has crept in | You can read the sentence once and say back who does what, how often, with nothing in it that belongs in a formula, and no single team can move it except by moving an input |
| 2 | Validation is honest | The claim that this metric predicts retention or revenue rests on nothing: no evidence, no plan to get any | Evidence is named, but it is another product’s data, or the plan to validate has no date and no owner | You can point to a correlation run against this product’s own data, or to a dated, owned plan to run one |
| 3 | Fits in one sentence | The definition takes more than one sentence, or leans on a term only the people who wrote it would recognize | One sentence, but it would need an example or a follow-up question before an outsider could use it | One sentence, plain enough that you can picture someone in sales or support repeating it back correctly, unprompted |
| 4 | Inputs are real levers | A row names a team with no measure, or restates the North Star itself under another label | Every row names a measure and an owner, but at least one owner would not actually notice it slip | For each row, you can name who would notice it slip this week and say in one sentence why moving it should move the North Star |
| 5 | Measures, not mirrors | The section could be lifted straight out of the product strategy with no change | It names the strategy and the vision, but the business outcome is vague, or written as if the metric were that outcome | You can say which strategy this number tests, and name the outcome it predicts without being that outcome |
| 6 | Trigger, not a date | The answer is a calendar entry, with nobody named to notice anything | An event is named, but nobody is named as responsible for noticing it | A specific shift is named, a specific person owns watching for it, and it is kept distinct from a target the number merely misses |
| 7 | Filters a real decision | Almost any project in flight could claim to serve this document; nothing here would rule one out | The document could rule something out in principle, but nobody has actually tried asking it of a real, named project | You can name a real project and show, from this document alone, whether it moves the North Star or not |
| 8 | Candidates were real (full) | Every rejected candidate is a straw metric nobody was proposing | A real candidate appears, but the reason it lost would equally disqualify the metric that was kept | At least one candidate could pass a casual glance, and its rejection names the specific test it failed |
| 9 | Guardrails earn their place (full) | A guardrail restates the North Star itself, or names a check with no stated harm | Each guardrail names a real measure, but for at least one you could not say what its decline would actually mean | Each guardrail could fall while the North Star climbs, you can state what that fall would mean, and none is the North Star’s own inverse |
| 10 | Owns its blind spots (full) | The section claims the metric has no real blind spots | A blind spot is named, but nothing elsewhere in the document catches the gap it describes | A blind spot is named before the metric ships, and a guardrail or an input exists in this document because of it |
| 11 | History survives the change (full) | An earlier definition is missing from a North Star that has in fact changed, and there is no “first definition” entry either | The new definition is recorded, but the old one, or the reason for the change, is missing | The old definition sits beside the new one, dated, with the shift in product, market, or strategy that caused it, or a dated “first definition” entry if there is no earlier one |
Which rows apply to what.
| Document | Rows | Maximum | Score against |
|---|---|---|---|
| full | all 11 | 22 | 15 |
| lean | 1-7 | 14 | 10 |
Rows 8 to 11 are scored only against full. The lean variant ships no Candidates Considered, Guardrails, What It Does Not Tell Us, or Definition History section, so grading it on those rows would penalise the choice of size rather than the quality of the document.
Failure signals to look for in the draft
Section titled “Failure signals to look for in the draft”Nothing here ranks against anything else; read it as a list, not an order.
- A metric that can only rise. Total registered users, cumulative downloads, and pageviews only go up, and a metric that can never fall cannot warn you when something is wrong. Test it with the doubling question: if this number doubled next month, would the business clearly be healthier? If not, it is a vanity metric wearing a North Star’s clothes.
- A metric the team can move directly. See the first test above. If your own team’s effort alone can push the number, it is a lever, not the thing the levers are supposed to move.
- Revenue, or any lagging measure, standing in for the metric. Every source that addresses this agrees on direction: revenue tells you what already happened, not what is about to.
- Gaming it once it becomes a target. Named independently, under different words, by a reference source asking how to keep the number from being gamed, and by a practitioner’s essay, South Star Metrics, about exactly this failure.
- Letting departments run their own, uncoordinated North Stars. One practitioner’s own warning is that this compounds as a company grows; nobody is left who can say which number the company is actually steering by.
- Silently rewriting the definition instead of recording the change. The discipline a glossary entry states directly: preserve the old definition and explain the break in the series rather than rewriting history.
- A composite score standing in for a single metric. Composite scores with several weighted inputs fail the one-sentence test even when they are, on paper, more complete.
- Mistaking it for a goal-setting method. Amplitude’s own playbook names this directly as what the Framework is not, and says so right next to the warning that it can still be a foundation goals are built on.
Before you share it
Section titled “Before you share it”- You can say the definition back in one sentence, and so could someone outside the team.
- Nothing in the document is a formula, a data source, a refresh schedule, a target, or a threshold.
- Every input names an owner who would actually notice it slip.
- The revisit trigger names an event and a person, never only a date.
- If this is the full variant: Candidates Considered, Guardrails, What It Does Not Tell Us, and Definition History are filled with real content, not left at “N/A” by default.
- You have deleted every HTML comment before sharing it.
The artifacts
Section titled “The artifacts”north-star-metrics_template-lean.md · ~3,850 tokens
---title: "{{product_name}} North Star Metric"product: "{{product_name}}"owner: "{{who_owns_this_metric}}"status: "{{draft_or_agreed}}"last_updated: "{{date}}"doc_type: north-star-metricssize: leansource_template: north-star-metricssource_template_version: 0.1.0---
<!--LEAN NORTH STAR METRIC. The shortest form that still names the metric, the case for it, its inputmetrics, what it serves, and the signal that would bring you back to it. Use it to settle what yourproduct's single measure of customer value is and why, before anyone asks what was rejected or whatwould change it. To grow it into a form that also names rejected candidates, standing guardrails, themetric's own blind spots, and a change log, see north-star-metrics_template-full.md. ADD sectionsafter When We Revisit It; never rename or reorder the five below, because the full variant is a strictordered superset of this one.
ADMISSION, AND WHAT IT MEANS FOR THIS TEMPLATE. This type is admitted on one publisher, Amplitude,across two documents that both reserve all rights. Every field name, section name and layout choicebelow is this library's own wording, never Amplitude's worksheet, its checklist text, or its sidebarheuristic. Where a sentence below is attributed to Amplitude or another source, it is reported as thatsource's own words, never offered as this template's own field label.
WHAT A NORTH STAR METRIC IS, AND IS NOT. It names the single measure of customer value that shows aproduct's strategy is working, states the case for it, and names the inputs a team can actually move.It is NOT the product vision (that names the destination; this names the number that shows thestrategy is getting you there). It is NOT the product strategy (that chooses which problems to solve;in this library's chain, drawn from sources that describe a North Star as measuring the currentstrategy, the North Star comes after it; one source orders the two the other way, andnorth-star-metrics_companion.md section 6 names both camps). It is NOT OKRs (those reset each period;a North Star persists and is revisited on a trigger, not a calendar). And it is NOT kpi-dashboard(that carries the formula, the data source, the refresh schedule, the target and the threshold; thisdocument carries one definition and the case for choosing it). See north-star-metrics_companion.mdsection 8 for the fuller set of boundaries.
WHAT STAYS OUT, AT EVERY SIZE. No formula, no data source, no refresh schedule, no target, nothreshold, no dashboard layout, and no tree of metrics below the inputs. All of that belongs tokpi-dashboard, except the metric tree, which the catalog lists as its own document type with the NorthStar at its top node. If you find yourself writing a calculation here, stop and open kpi-dashboardinstead.
ONE NORTH STAR, PER PRODUCT OR PER PRODUCT LINE. The default is one North Star metric for the wholeproduct. The named exception is a company with genuinely separate product lines, each with its owncustomers and its own profit and loss; that company may need one North Star per line, not one per teamor one per feature. A direct competitor to the admitting publisher disputes the one-metric premise as"mostly rhetorical"; this bundle follows the one-per-product rule with the product-line exception. Seenorth-star-metrics_companion.md section 6.
HOW TO FILL THIS IN1. Read the comment under each heading: WHAT it wants, WHY it matters (with a pointer into north-star-metrics_companion.md), guiding questions to ASK, a GOOD and a WEAK example, and the TRAP to avoid. For the table, PRIORITY explains the ordering rule and ROW HINT says what a good row contains.2. Replace each {{placeholder}} with your content. Name no formula, data source, refresh schedule, target, or threshold anywhere in this document; those belong to kpi-dashboard.3. If a section does not apply, write "N/A" and one line of why, rather than deleting it.4. Before you share it: self-grade against north-star-metrics_guide.md, then DELETE every HTML comment. They are guidance, not content.-->
# {{product_name}} North Star Metric
## The North Star
<!-- WHAT The metric's name and its definition in one statement: the behavior it counts, who does it, and over what time window. The product it covers, if your company runs more than one. WHY Cut the SaaS states the window directly: "Pick a metric that counts that action across a sensible time window, like weekly." Ballpark adds the bar for the definition itself: "Prefer a definition people can explain, inspect and reproduce." Amplitude's own fill-in exercise, reported here and not reproduced as this template's own field, asks for the metric's name and a one-sentence definition; this section asks for the same two things in the library's own wording. Deep dive: north-star-metrics_companion.md section 3 (The North Star). ASK What behavior does this metric count? Who performs it, and over what window, weekly or some other sensible period? Could someone outside your team read the definition and reproduce the count themselves? Does your company run more than one product, and if so, which one does this North Star belong to? GOOD "Our North Star is weekly closed expense reports: the count of expense reports that reach Approved or Reimbursed status, per active finance team, each week, with no manual nudge from a success manager." (names the behavior, the actor, and the window, and leaves the calculation out) WEAK "Our North Star is engagement." (no behavior, no actor, no window; nobody could reproduce a count from this) TRAP Writing the calculation instead of the definition. A column for exclusions, a query snippet, or a refresh schedule belongs in kpi-dashboard's Metric Definitions section, not here. -->
{{north_star_definition}}
- Product this North Star belongs to, if your company has more than one: {{product_scope}}
## Why This Metric
<!-- WHAT The case for this metric over another: the customer value it expresses, why it leads revenue rather than reporting it, why it is not a vanity metric, and what shows it predicts retention or revenue in this product's own data. WHY Amplitude's own checklist asks, among other things, "Does it express customer value?" and "Is it a leading indicator of success?" A second Amplitude article adds the instruction to check the claim against your own evidence: "Validate that it predicts retention and revenue in your own data." Koji states where that evidence has to come from: "You cannot write this sentence by staring at a dashboard." Deep dive: north-star-metrics_companion.md section 3 (Why This Metric). ASK What does this metric say about the value a customer actually gets? Why does it move before revenue rather than after? Could it rise forever even if the business got no healthier, the doubling test one source proposes? What shows, in your own data or your own customer conversations, that this metric predicts retention or revenue, and if nothing does yet, how will you find out? GOOD "Closed expense reports measures the moment a finance team gets what it pays us for: an expense submitted, reviewed and settled without chasing anyone. It is not revenue, which only tells us the deal closed, not whether the team is getting value every week. We have not yet run the correlation against renewal; that test is scheduled for next quarter." (states the value, the leading-versus-lagging case, and labels the missing validation honestly) WEAK "This metric matters because it's the number that matters most." (restates the question; names no customer value, no leading-indicator case, no validation) TRAP Treating "it predicts revenue" as self-evident. The admitting source's own instruction is to validate the claim in your own data, not to assume a metric is a leading indicator because it sounds like one. -->
- The customer value case: {{customer_value_case}}- Why it leads rather than lags revenue: {{leading_indicator_case}}- Why it is not a vanity metric: {{not_vanity_metric_case}}- What shows it predicts retention or revenue in your own data, or how you will find out: {{validation_evidence}}
## Input Metrics
<!-- WHAT Three to five measures a team can actually move, that combine to drive the North Star, each with a named owner. WHY Amplitude defines inputs as "a set of three to five influential, complementary factors that you believe most directly affect your North Star Metric and that your team can directly influence through your product." Basedash states the ownership rule: "Each input is something a specific team can own and affect directly." A reference source goes further, naming six to ten input metrics across a customer journey with each team owning one to three of them; this bundle keeps Amplitude's three to five as its default and reports the larger count as that one source's own. Amplitude's own breadth, depth, frequency and efficiency heuristic for finding inputs is reported here as one vendor's own, never prescribed; a competing vendor structures inputs as a hierarchy instead, and this bundle prescribes neither. Deep dive: north-star-metrics_companion.md section 3 (Input Metrics), section 6 (How input metrics are organized). ASK What three to five measures, if your team moved them, would move the North Star? Who owns each one, specifically enough that they would notice it slipping? Could the North Star move without any of these moving, in which case is something missing? Is any row really the North Star itself, counted a different way? PRIORITY Three to five is this bundle's default. One reference source allows six to ten across a longer customer journey, with one to three per team. Past five rows, every one still needs an owner who can actually move it, or it has not earned its place. ROW HINT A good row names one measurable behavior, one owning team, and why moving that behavior should move the North Star. A weak row names a team with no measure, or restates the North Star's own definition under a different label. GOOD | Receipt-capture completion rate | Mobile team | A report cannot close until its receipts are attached; this is the most common place a report stalls | WEAK | Customer happiness | Everyone | | TRAP Adding a row because a team wants one, not because moving it would move the North Star. A row with no real lever behind it is the same failure Amplitude's own playbook names for the North Star itself: insisting on more than real boundaries support. -->
| Input Metric | Owning Team | Why It Drives the North Star ||---|---|---|| {{input_metric_name}} | {{input_metric_owner}} | {{input_metric_rationale}} |
## What It Serves
<!-- WHAT The product strategy this metric measures, the vision that strategy serves, and the business outcome this metric is expected to predict. WHY Amplitude names one part of a good North Star as "a metric that serves as a key measure of your current product strategy." This library places the North Star after the strategy it measures; one practitioner orders the two the other way, putting the North Star first. Both camps are named in north-star-metrics_companion.md section 6 (The chain position is contested); this section states this library's position and asks you to name your strategy, not to settle the debate. Deep dive: north-star-metrics_companion.md section 3 (What It Serves). ASK Which product strategy does this number show is working? How does that strategy serve the product vision? What business outcome, revenue, retention, or expansion, is this metric expected to predict, and does it correlate with that outcome without being that outcome? GOOD "Closed expense reports measures whether our strategy, becoming the system finance teams trust enough to stop keeping a shadow spreadsheet, is working. It serves the vision of a finance team that never reconciles a paper receipt by hand. It is expected to predict renewal, not replace it as the renewal number itself." WEAK "This metric supports our overall goals." (names no strategy, no vision, no outcome) TRAP Copying the strategy's own language in here unexamined. If this section could be lifted straight out of the strategy document with no change, ask whether the metric actually measures the strategy or is simply repeating it. -->
- The product strategy this metric measures: {{strategy_it_measures}}- The product vision that strategy serves: {{vision_it_serves}}- The business outcome it is expected to predict, and how it differs from that outcome: {{business_outcome_expected}}
## When We Revisit It
<!-- WHAT The signal that would bring the team back to this definition, not a calendar date. WHY Amplitude's own playbook asks, "What will be your signal to revisit this metric?" A reference source states the trigger from the other side: "Only change the NSM when strategy or product fundamentally shifts." Spotify's experimentation glossary names the stakes: "Changing the north star is a significant organizational decision." Sources disagree sharply on cadence, by stage, by a flat yearly rule, or on a fixed quarterly calendar; this section asks for the trigger because no source read agrees on a number. Deep dive: north-star-metrics_companion.md section 3 (When We Revisit It), section 6 (How often it changes). ASK What change in the product, the market, or the strategy would make you look at this definition again? Who notices that signal first? What would have to be true for the metric itself, not just its target, to change? GOOD "We revisit this definition if we start selling to teams that approve travel budgets rather than individual expense reports; closed expense reports would no longer describe the value those teams get. The head of product owns watching for that shift." WEAK "We'll review this every quarter." (a calendar entry, naming no event and nobody responsible for noticing one) TRAP Confusing a trigger with a target miss. A quarter where the number does not move is a signal to investigate the input metrics, not, on its own, a reason to redefine the North Star itself. -->
- The signal that would bring you back to this definition, not a date: {{revisit_trigger}}- Who is responsible for noticing that signal: {{revisit_owner}}north-star-metrics_template-full.md · ~6,250 tokens
---title: "{{product_name}} North Star Metric"product: "{{product_name}}"owner: "{{who_owns_this_metric}}"status: "{{draft_or_agreed}}"last_updated: "{{date}}"doc_type: north-star-metricssize: fullsource_template: north-star-metricssource_template_version: 0.1.0---
<!--FULL NORTH STAR METRIC. Every section of north-star-metrics_template-lean.md, in the same order, plusthe four a North Star needs once it has to survive a reader asking what was rejected, what must notget worse while it climbs, what it cannot show, and what the definition looked like before: CandidatesConsidered, Guardrails, What It Does Not Tell Us, and Definition History, appended after When WeRevisit It.
USE FULL WHEN the North Star has already faced, or will soon face, real scrutiny: named alternativeswere seriously compared, a stakeholder will ask "why not X," or the definition has already changedonce and the change needs a record. Use lean (north-star-metrics_template-lean.md) for a first pass,or where nobody is likely to ask what was rejected.
ADMISSION, AND WHAT IT MEANS FOR THIS TEMPLATE. This type is admitted on one publisher, Amplitude,across two documents that both reserve all rights. Every field name, section name and layout choicebelow is this library's own wording, never Amplitude's worksheet, its checklist text, or its sidebarheuristic. Where a sentence below is attributed to Amplitude or another source, it is reported as thatsource's own words, never offered as this template's own field label.
GUARDRAILS IS THE ONE SECTION THAT DOES NOT REST ON THE ADMITTING PUBLISHER. Amplitude's own twodocuments never use "guardrail" for a measure tracked beside the North Star; their one use of the worddescribes anti-examples brainstormed during definition, not a field in the finished document. Thissection rests on three other sources instead, named where the section itself appears below. A teamfilling only the lean variant is not skipping an Amplitude step by leaving Guardrails out.
WHAT A NORTH STAR METRIC IS, AND IS NOT. It names the single measure of customer value that shows aproduct's strategy is working, states the case for it, and names the inputs a team can actually move.It is NOT the product vision (that names the destination; this names the number that shows thestrategy is getting you there). It is NOT the product strategy (that chooses which problems to solve;in this library's chain, drawn from sources that describe a North Star as measuring the currentstrategy, the North Star comes after it; one source orders the two the other way, andnorth-star-metrics_companion.md section 6 names both camps). It is NOT OKRs (those reset each period;a North Star persists and is revisited on a trigger, not a calendar). And it is NOT kpi-dashboard(that carries the formula, the data source, the refresh schedule, the target and the threshold; thisdocument carries one definition and the case for choosing it). See north-star-metrics_companion.mdsection 8 for the fuller set of boundaries.
WHAT STAYS OUT, AT EVERY SIZE. No formula, no data source, no refresh schedule, no target, nothreshold, no dashboard layout, and no tree of metrics below the inputs. All of that belongs tokpi-dashboard, except the metric tree, which the catalog lists as its own document type with the NorthStar at its top node. If you find yourself writing a calculation here, stop and open kpi-dashboardinstead.
ONE NORTH STAR, PER PRODUCT OR PER PRODUCT LINE. The default is one North Star metric for the wholeproduct. The named exception is a company with genuinely separate product lines, each with its owncustomers and its own profit and loss; that company may need one North Star per line, not one per teamor one per feature. A direct competitor to the admitting publisher disputes the one-metric premise as"mostly rhetorical"; this bundle follows the one-per-product rule with the product-line exception. Seenorth-star-metrics_companion.md section 6.
HOW TO FILL THIS IN1. Read the comment under each heading: WHAT it wants, WHY it matters (with a pointer into north-star-metrics_companion.md), guiding questions to ASK, a GOOD and a WEAK example, and the TRAP to avoid. For the tables, PRIORITY explains the ordering rule and ROW HINT says what a good row contains.2. Replace each {{placeholder}} with your content. Name no formula, data source, refresh schedule, target, or threshold anywhere in this document; those belong to kpi-dashboard.3. If a section does not apply, write "N/A" and one line of why, rather than deleting it.4. Before you share it: self-grade against north-star-metrics_guide.md, then DELETE every HTML comment. They are guidance, not content.-->
# {{product_name}} North Star Metric
## The North Star
<!-- WHAT The metric's name and its definition in one statement: the behavior it counts, who does it, and over what time window. The product it covers, if your company runs more than one. WHY Cut the SaaS states the window directly: "Pick a metric that counts that action across a sensible time window, like weekly." Ballpark adds the bar for the definition itself: "Prefer a definition people can explain, inspect and reproduce." Amplitude's own fill-in exercise, reported here and not reproduced as this template's own field, asks for the metric's name and a one-sentence definition; this section asks for the same two things in the library's own wording. Deep dive: north-star-metrics_companion.md section 3 (The North Star). ASK What behavior does this metric count? Who performs it, and over what window, weekly or some other sensible period? Could someone outside your team read the definition and reproduce the count themselves? Does your company run more than one product, and if so, which one does this North Star belong to? GOOD "Our North Star is weekly closed expense reports: the count of expense reports that reach Approved or Reimbursed status, per active finance team, each week, with no manual nudge from a success manager." (names the behavior, the actor, and the window, and leaves the calculation out) WEAK "Our North Star is engagement." (no behavior, no actor, no window; nobody could reproduce a count from this) TRAP Writing the calculation instead of the definition. A column for exclusions, a query snippet, or a refresh schedule belongs in kpi-dashboard's Metric Definitions section, not here. -->
{{north_star_definition}}
- Product this North Star belongs to, if your company has more than one: {{product_scope}}
## Why This Metric
<!-- WHAT The case for this metric over another: the customer value it expresses, why it leads revenue rather than reporting it, why it is not a vanity metric, and what shows it predicts retention or revenue in this product's own data. WHY Amplitude's own checklist asks, among other things, "Does it express customer value?" and "Is it a leading indicator of success?" A second Amplitude article adds the instruction to check the claim against your own evidence: "Validate that it predicts retention and revenue in your own data." Koji states where that evidence has to come from: "You cannot write this sentence by staring at a dashboard." Deep dive: north-star-metrics_companion.md section 3 (Why This Metric). ASK What does this metric say about the value a customer actually gets? Why does it move before revenue rather than after? Could it rise forever even if the business got no healthier, the doubling test one source proposes? What shows, in your own data or your own customer conversations, that this metric predicts retention or revenue, and if nothing does yet, how will you find out? GOOD "Closed expense reports measures the moment a finance team gets what it pays us for: an expense submitted, reviewed and settled without chasing anyone. It is not revenue, which only tells us the deal closed, not whether the team is getting value every week. We have not yet run the correlation against renewal; that test is scheduled for next quarter." (states the value, the leading-versus-lagging case, and labels the missing validation honestly) WEAK "This metric matters because it's the number that matters most." (restates the question; names no customer value, no leading-indicator case, no validation) TRAP Treating "it predicts revenue" as self-evident. The admitting source's own instruction is to validate the claim in your own data, not to assume a metric is a leading indicator because it sounds like one. -->
- The customer value case: {{customer_value_case}}- Why it leads rather than lags revenue: {{leading_indicator_case}}- Why it is not a vanity metric: {{not_vanity_metric_case}}- What shows it predicts retention or revenue in your own data, or how you will find out: {{validation_evidence}}
## Input Metrics
<!-- WHAT Three to five measures a team can actually move, that combine to drive the North Star, each with a named owner. WHY Amplitude defines inputs as "a set of three to five influential, complementary factors that you believe most directly affect your North Star Metric and that your team can directly influence through your product." Basedash states the ownership rule: "Each input is something a specific team can own and affect directly." A reference source goes further, naming six to ten input metrics across a customer journey with each team owning one to three of them; this bundle keeps Amplitude's three to five as its default and reports the larger count as that one source's own. Amplitude's own breadth, depth, frequency and efficiency heuristic for finding inputs is reported here as one vendor's own, never prescribed; a competing vendor structures inputs as a hierarchy instead, and this bundle prescribes neither. Deep dive: north-star-metrics_companion.md section 3 (Input Metrics), section 6 (How input metrics are organized). ASK What three to five measures, if your team moved them, would move the North Star? Who owns each one, specifically enough that they would notice it slipping? Could the North Star move without any of these moving, in which case is something missing? Is any row really the North Star itself, counted a different way? PRIORITY Three to five is this bundle's default. One reference source allows six to ten across a longer customer journey, with one to three per team. Past five rows, every one still needs an owner who can actually move it, or it has not earned its place. ROW HINT A good row names one measurable behavior, one owning team, and why moving that behavior should move the North Star. A weak row names a team with no measure, or restates the North Star's own definition under a different label. GOOD | Receipt-capture completion rate | Mobile team | A report cannot close until its receipts are attached; this is the most common place a report stalls | WEAK | Customer happiness | Everyone | | TRAP Adding a row because a team wants one, not because moving it would move the North Star. A row with no real lever behind it is the same failure Amplitude's own playbook names for the North Star itself: insisting on more than real boundaries support. -->
| Input Metric | Owning Team | Why It Drives the North Star ||---|---|---|| {{input_metric_name}} | {{input_metric_owner}} | {{input_metric_rationale}} |
## What It Serves
<!-- WHAT The product strategy this metric measures, the vision that strategy serves, and the business outcome this metric is expected to predict. WHY Amplitude names one part of a good North Star as "a metric that serves as a key measure of your current product strategy." This library places the North Star after the strategy it measures; one practitioner orders the two the other way, putting the North Star first. Both camps are named in north-star-metrics_companion.md section 6 (The chain position is contested); this section states this library's position and asks you to name your strategy, not to settle the debate. Deep dive: north-star-metrics_companion.md section 3 (What It Serves). ASK Which product strategy does this number show is working? How does that strategy serve the product vision? What business outcome, revenue, retention, or expansion, is this metric expected to predict, and does it correlate with that outcome without being that outcome? GOOD "Closed expense reports measures whether our strategy, becoming the system finance teams trust enough to stop keeping a shadow spreadsheet, is working. It serves the vision of a finance team that never reconciles a paper receipt by hand. It is expected to predict renewal, not replace it as the renewal number itself." WEAK "This metric supports our overall goals." (names no strategy, no vision, no outcome) TRAP Copying the strategy's own language in here unexamined. If this section could be lifted straight out of the strategy document with no change, ask whether the metric actually measures the strategy or is simply repeating it. -->
- The product strategy this metric measures: {{strategy_it_measures}}- The product vision that strategy serves: {{vision_it_serves}}- The business outcome it is expected to predict, and how it differs from that outcome: {{business_outcome_expected}}
## When We Revisit It
<!-- WHAT The signal that would bring the team back to this definition, not a calendar date. WHY Amplitude's own playbook asks, "What will be your signal to revisit this metric?" A reference source states the trigger from the other side: "Only change the NSM when strategy or product fundamentally shifts." Spotify's experimentation glossary names the stakes: "Changing the north star is a significant organizational decision." Sources disagree sharply on cadence, by stage, by a flat yearly rule, or on a fixed quarterly calendar; this section asks for the trigger because no source read agrees on a number. Deep dive: north-star-metrics_companion.md section 3 (When We Revisit It), section 6 (How often it changes). ASK What change in the product, the market, or the strategy would make you look at this definition again? Who notices that signal first? What would have to be true for the metric itself, not just its target, to change? GOOD "We revisit this definition if we start selling to teams that approve travel budgets rather than individual expense reports; closed expense reports would no longer describe the value those teams get. The head of product owns watching for that shift." WEAK "We'll review this every quarter." (a calendar entry, naming no event and nobody responsible for noticing one) TRAP Confusing a trigger with a target miss. A quarter where the number does not move is a signal to investigate the input metrics, not, on its own, a reason to redefine the North Star itself. -->
- The signal that would bring you back to this definition, not a date: {{revisit_trigger}}- Who is responsible for noticing that signal: {{revisit_owner}}
## Candidates Considered
<!-- WHAT The other metrics genuinely on the table, and the specific test each one failed. WHY Amplitude's own worked example describes a company whose rejected candidates "masked the real drivers of user retention and customer lifetime value"; a second Amplitude article reports a team that kept its chosen metric over alternatives "because your data shows that customers value" what the metric actually counts. Basedash names the test for the most common failure mode: "Total registered users, cumulative downloads, and pageviews only go up. A metric that can never fall cannot warn you when something is wrong." Cut the SaaS states the same test from a sibling source: "Total signups, page views, and registered users all fail that test." Deep dive: north-star-metrics_companion.md section 3 (Candidates Considered). ASK What else was seriously proposed, not only the obvious vanity metric? What specific test did each one fail, can it only rise, is it revenue in disguise, does it mask what is actually driving retention? Who proposed the one you kept, and what changed their mind about the others? PRIORITY Include at least one candidate that could pass a casual glance, not only a straw metric nobody was proposing. A table of candidates nobody believed in proves nothing about the choice you actually made. ROW HINT A good row names a metric that was genuinely considered and the specific reason it lost, tied to one of the tests above. A weak row rejects a metric nobody proposed, or gives a reason that would equally reject the metric you kept. GOOD | Total expense reports submitted | Can only rise; a team could submit reports and never close one and this number would still climb | WEAK | Number of users | Not good enough | TRAP Rejecting only metrics nobody wanted. If every candidate here was obviously bad, the real contest, the one between your North Star and its closest real competitor, is still hiding. -->
| Candidate Metric | Why It Was Rejected ||---|---|| {{candidate_metric}} | {{why_rejected}} |
## Guardrails
<!-- WHAT The measures that must not get worse while the North Star climbs. WHY This section does not rest on the admitting publisher's own canon, and says so plainly. Amplitude's two documents never use "guardrail" for a measure tracked beside the North Star; their one use of the word describes anti-examples brainstormed during definition, not a field in the finished document. This section rests on three other sources instead: Cut the SaaS, "The fix is pairing your north star with guardrail metrics like retention and gross margin"; Ballpark, "Pair the shared measure with checks for harm or declining quality"; and a reference source that names the shape as "Quality and economics limits to prevent 'hitting the target but missing the point.'" Two further sources scope the same word more narrowly, to a single experiment's safety check; this section follows the standing-companion sense instead. Deep dive: north-star-metrics_companion.md section 3 (Guardrails), section 6 (Whether guardrails belong beside the North Star). ASK What could get worse while this North Star climbs, quality, margin, or trust, that nobody would notice without a named check? What specific decline would each guardrail catch? Is any guardrail here really just the North Star's inverse, which would not be a second check at all? PRIORITY Keep this short enough that someone could name every row without looking it up. Two or three is this library's own rule of thumb; no source read fixes a count for a standing guardrail list. ROW HINT A good row names a measure that could decline even as the North Star rises, and states plainly what that decline would mean. A weak row restates the North Star itself, or names a check with no stated harm. GOOD | Policy-violation rate on closed reports | Closed reports could rise because approvers are waving through anything, not because the process got easier | WEAK | Customer satisfaction | | TRAP Pairing the North Star with so many guardrails that nobody can hold all of them in mind. Two or three, chosen for the specific way this North Star could be gamed, do more than a long checklist. -->
| Guardrail Metric | What Decline Would Mean ||---|---|| {{guardrail_metric}} | {{what_decline_means}} |
## What It Does Not Tell Us
<!-- WHAT The metric's blind spots, stated honestly: what it cannot show, and how it could be gamed. WHY Amplitude's own playbook asks, "What doesn't this metric tell you? Are any of those questions or uncertainties more important to address?" A reference source asks the sharper version of the same question: "How do we keep the NSM from being gamed?" Spotify's experimentation glossary names a structural limit worth stating here rather than assuming it: "North star metrics tend to be high-level and slow-moving." Deep dive: north-star-metrics_companion.md section 3 (What It Does Not Tell Us). ASK What question would someone reasonably ask that this number cannot answer on its own? How slow is it to move, and what would a team need to watch instead while waiting for it to show a change? How could someone make this number go up without the customer getting any more value, and what would that look like if it happened? GOOD "Closed expense reports cannot tell us whether a report closed quickly or sat for three weeks before a tired approver waved it through; that is why the policy-violation guardrail exists. It also will not move visibly inside a single sprint, so sprint-level decisions use the input metrics instead." WEAK "This metric has no real blind spots." (the admitting source's own question, "what doesn't this metric tell you," assumes there are some) TRAP Treating this section as a formality filled in after the fact. A blind spot named here before the metric ships is a question the team can watch for; one discovered after launch is usually discovered by someone gaming it. -->
- What this metric cannot show on its own: {{known_blind_spots}}- How it could be gamed, and what that would look like: {{gaming_risk}}
## Definition History
<!-- WHAT Every definition this North Star has carried, oldest first, with the date and the reason it changed. WHY Ballpark states the discipline directly: "Preserve the old definition and explain the break in the series rather than silently rewriting history." A competing vendor states the same discipline as a governance obligation: "Every metric should have clear documentation - how it's defined, calculated, and where the data comes from." Amplitude's own first-person account of replacing one North Star with another, with its reasons, is this bundle's worked illustration of what one entry in this history looks like. Deep dive: north-star-metrics_companion.md section 3 (Definition History). ASK Has this North Star had an earlier definition? What was it, word for word, not paraphrased? What changed in the product, market, or strategy that made the old definition stop fitting? Is the old entry still here, or did it get quietly deleted? PRIORITY List entries oldest first. Never delete a retired definition; a definition history with one entry removed is not a history. ROW HINT A good row states the old definition and the new one side by side, the date, and the shift in product, market, or strategy that caused it. A weak row states only the new definition, or gives "it fits better now" as the whole reason. GOOD | 2025-11-01 | Total expense reports submitted | Closed expense reports | The old definition rose even as approvers complained reports were piling up unresolved; it measured activity, not value received | WEAK | 2025-11-01 | | Closed expense reports | We decided to change it | TRAP Silently rewriting the old definition instead of logging the change. If this is the first definition this North Star has ever carried, write "first definition" and today's date rather than leaving the table empty. -->
| Date | Previous Definition | New Definition | Why It Changed ||---|---|---|---|| {{history_date}} | {{previous_definition}} | {{new_definition}} | {{why_it_changed}} |---title: "Acme Analytics North Star Metric"product: "Acme Analytics"owner: "Dana Okoro, VP Product"status: "agreed"last_updated: "2026-02-04"doc_type: north-star-metricssize: fullsource_template: north-star-metricssource_template_version: 0.1.0---
> **Worked example.** A filled `north-star-metrics`, full variant, for the same Acme Analytics product used> throughout this library's examples. Per [ADR 0070](../../docs/internal/decisions/0070-north-star-metrics-joins-strategy-docs-as-a-fifth-member.md),> this type sits third in the `strategy-docs` chain: it is dated 2026-02-04, one week after the> [FY26 product strategy](../product-strategy/product-strategy_example.md) (2026-01-28) it measures, and it> will sit just ahead of the product roadmap still to come. Dana Okoro, VP Product, owns the vision, the> strategy, this document, and the roadmap that follows it, which is why the four speak in one voice. Nothing> below carries a formula, a data source, a refresh schedule, a target, or a threshold; those belong to a KPI> dashboard, not to this document. *(2026-10-20: since [ADR 0076](../../docs/internal/decisions/0076-positioning-statement-joins-strategy-docs-as-a-sixth-member.md) placed the positioning statement third, this type sits fourth in the chain. Nothing else here changes.)*>> All figures are illustrative. They are internally consistent with the other examples in this library and> are not drawn from a real company. Any figure this document shares with an earlier example, the 31 percent> baseline and the second-view measure above all, repeats that example's own number exactly.
# Acme Analytics North Star Metric
## The North Star
Our North Star is **Weekly Self-Served Outcomes**: the count, across every active Acme Analytics account, ofa working session in which someone outside the analytics team opens a report, gets an answer to the questionthat made them open it, and acts on that answer in the same session, by exporting it, sharing it, annotatingit, or running a follow-on query, without asking Dev or the analytics team first. Counted weekly, across thewhole account base, not only new accounts. The exact event definitions, the data source, and the weeklytarget this North Star is held to live on the dashboard our data team maintains, not here.
- Product this North Star belongs to, if your company has more than one: This North Star covers Acme Analytics only, not the rest of Acme's portfolio.
## Why This Metric
- The customer value case: The [2026-01-14 product vision](../product-vision/product-vision_example.md) defines the value we are trying to create as the end of the waiting period between noticing something and understanding it, stated as someone who "can answer a question about their own operations in the time it takes to ask it out loud." A view someone opens and never acts on has not closed that gap; this count only moves when the gap actually closes, in the same session the question was asked.- Why it leads rather than lags revenue: It measures behavior inside the product this week, not a renewal decision that lands months later. An account can produce a self-served outcome today whether or not it renews next year, which is exactly why it has to move before self-serve net revenue retention does, not after.- Why it is not a vanity metric: It is a count of actions, not of accounts logged in or pages viewed, and it can fall. An account that stops acting on what it sees, even while it keeps opening reports, produces fewer outcomes, not more; a metric built from raw usage would keep climbing through exactly that failure.- What shows it predicts retention or revenue in your own data, or how you will find out: Not yet validated. We have not run this count against a real account's renewal outcome, because the count itself does not exist in production yet. Once the Reporting team's event instrumentation ships in Q2, Priya Nair's team will pull the first cohort of accounts with two full quarters of this count against their Q4 renewal decision and report whether the two move together.
## Input Metrics
| Input Metric | Owning Team | Why It Drives the North Star ||---|---|---|| Share of new accounts that build a second view within 14 days of signup, from a 31 percent baseline in Q4 FY25 | Reporting team (Priya Nair) | Already the number every team watches weekly on the shared dashboard, and the strategy's own leading indicator, so near-term planning ladders to this number before it ladders to anything else; an account that never builds a second view has no second surface from which a self-served outcome could ever occur, and the North Star above shows whether that first activation turns into repeated value or stops there || Per-vertical modelling-default coverage: share of new accounts whose first view is built from a shipped default rather than a blank canvas | Platform team | Question-first entry can only resolve a question to something real if a modelled default already exists for that account's vertical; an account with no default behind it has nothing to answer against || Question-first resolution rate: share of question-first queries that resolve to the right dataset and join without the user correcting them | Platform team | A query that resolves wrong produces a correction, not an answer, so it cannot become an outcome someone acts on in that same session |
## What It Serves
- The product strategy this metric measures: The FY26 strategy to make the product usable without data expertise, "rather than making expertise easier to acquire." Every one of its four coherent actions, the question-first entry point, the shipped modelling defaults, instrumenting the second view, and moving services capacity toward defaults, is aimed at producing more sessions where a non-expert gets and uses an answer, which is what this North Star counts.- The product vision that strategy serves: The 2029 vision of ending the wait between noticing something and understanding it, and of "Removing the analytics team as a bottleneck" on every other product decision. A self-served outcome is that vision happening once, in one session, for one person.- The business outcome it is expected to predict, and how it differs from that outcome: It is expected to predict both of the strategy's lagging measures, the FY26 "Time to Insight" company goal and self-serve net revenue retention. It differs from both: it is a weekly count of sessions, not a median time and not a renewal percentage, and it is designed to move before either of them does, not to restate either one in a different unit.
## When We Revisit It
- The signal that would bring you back to this definition, not a date: Two signals. First, the 2026-01-14 vision already names a leap of faith we are watching in the pilot: three of eleven pilot users still take our answers to Dev to check them before acting. If that share does not fall as familiarity with the product grows, the sessions this count calls "self-served" are not what the vision means by the word, and the definition needs to change, not just the number. Second, if Acme starts selling to a buyer who is required to route a decision through a data team for compliance reasons, acting without that team stops being possible for that account regardless of how the product works, and the definition would need to say so rather than silently undercount them.- Who is responsible for noticing that signal: Dana Okoro, VP Product.
## Candidates Considered
| Candidate Metric | Why It Was Rejected ||---|---|| The FY26 "Time to Insight" company goal | The strategy already calls this a lagging measure. A number that only tells you what happened after the fact cannot also be the number that predicts it. || Share of new accounts that build a second view within 14 days | Kept, but as Input 1, not as the North Star. It counts new accounts only, inside their first 14 days, so it goes silent for every account that passed day 14 months ago; it cannot show whether an account that activated last spring is still getting value every week. || Self-serve net revenue retention | A revenue number one layer removed from the product. It moves on a renewal cycle, not weekly, and a renewal can rise or fall for reasons that have nothing to do with whether this week's questions got answered. |
## Guardrails
| Guardrail Metric | What Decline Would Mean ||---|---|| Share of self-served outcomes that get taken to Dev afterward to double-check | If this climbs even as the North Star climbs, the sessions we are counting are not evidence of trust in the answer; the 2026-01-14 vision names exactly this failure as building "a faster way to create work for" the team we were trying to remove as a bottleneck. || Services-ticket requests to build or verify a view, filed by an account in the same week it produced a self-served outcome | If this rises alongside the North Star, some share of the outcomes we are counting were actually delivered by a person behind the scenes, not by the product acting self-served at all. |
## What It Does Not Tell Us
- What this metric cannot show on its own: Whether the question acted on mattered. A user who exports a chart confirming something they already half-knew counts exactly the same as a user who caught a real problem before it cost the business anything. A single account can also sit flat for a whole week without that meaning anything is wrong; a healthy account can go a week with no new question worth acting on, and one quiet week is not a signal the input metrics above do not already explain.- How it could be gamed, and what that would look like: A team under pressure to move the number could encourage exporting or annotating as a habit regardless of whether a decision actually followed, which would look like this count rising while the services-ticket guardrail above also rises, because the "self-served" action was decorative rather than load-bearing.
## Definition History
| Date | Previous Definition | New Definition | Why It Changed ||---|---|---|---|| 2026-02-04 | N/A, first definition | Weekly Self-Served Outcomes, as defined above | First definition. Recorded here, with today's date, so that any future change to this metric has to add a row rather than quietly rewriting this one. |Provenance
Section titled “Provenance”The reasoning, the history and every source, in the repository:
- Companion - the long-form argument: why these sections, where the sources disagree, and what the bundle refuses to claim
- History - what changed in this bundle, and when
- Research log - every source consulted, with what each one actually supports
- Catalog metadata - the machine-readable record this page is generated from
Catalog record: 9 sections across 1 format(s).