Choosing one north star metric for marketing

1. Start with the value customers actually receive

A north star metric is the main outcome used to align marketing decisions. It is not your only measure, a channel target or a claim that marketing controls the whole customer journey. Choose a meaningful customer action that connects to commercial health, then state where marketing's influence begins and ends.

For a hypothetical equipment repair business, completed paid repairs might be stronger than enquiry forms. Customers receive a working appliance, and the business earns revenue. However, technician availability also affects completions. Marketing and operations must therefore share the outcome rather than treating every movement as marketing performance.

Start with the sentence: ‘We help this customer achieve this outcome, and we observe that value when this event happens.’ If the event is merely seeing an advert, move further along the journey.

• Identify the customer, delivered value and observable event.
• Check that increasing the event supports the business model.
• Name the teams whose decisions can constrain the outcome.

2. Compare candidates instead of choosing the easiest dashboard number

Shortlist three candidates and assess each for customer value, commercial relevance, marketing influence, measurement reliability and reporting delay. Do not disguise judgement with a weighted score whose weights nobody can defend. Record the tradeoffs in plain language.

In a hypothetical software business, trial registrations are fast but can reward low-intent acquisition. Activated accounts are closer to experienced value, provided activation represents useful behaviour rather than a login. Retained paying accounts are commercially stronger but slower and more dependent on product quality.

Reject candidates with critical weaknesses rather than averaging those weaknesses away. A metric that cannot be deduplicated reliably is not rescued by being commercially attractive. Where no candidate is ready, use a clearly labelled provisional measure while fixing instrumentation.

• Value test: could this rise while customers receive less value?
• Action test: can the team name decisions that plausibly influence it?
• Trust test: can another analyst reproduce the number?
• Timing test: will feedback arrive soon enough to guide decisions?

3. Choose the right unit: count, rate or monetary value

Counts reward scale but can conceal deteriorating economics. Rates reveal efficiency but can improve when the business shrinks. Monetary measures reflect value differences, yet revenue alone ignores discounts, refunds and delivery costs. Choose deliberately rather than calling all three north stars.

A hypothetical retailer could increase conversion rate by stopping prospecting and serving only returning customers. The rate improves while new-customer volume falls. Net fulfilled orders may better express its immediate growth objective, with contribution margin and new-customer share as guardrails.

Where reliable cost data exists, contribution from fulfilled orders may be a stronger primary measure. Define exactly which costs are deducted. Avoid mixing orders, satisfaction and revenue into a custom index: a rising score can conceal a falling customer outcome.

• For a rate, specify both numerator and denominator.
• For money, define currency, tax treatment, refunds and costs.
• For counts, specify the unique entity and repeat-event rules.

4. Write a metric contract before setting a target

A metric needs an operational definition, not just a label. In a hypothetical consultancy, ‘qualified opportunities’ might mean unique buying organisations accepted by sales after confirming service fit, a named decision-maker and an agreed next meeting. A downloaded brochure would not qualify.

The contract should specify the reporting source, timezone, event date, exclusions, owner and correction policy. Decide whether an opportunity belongs to the week of first enquiry or sales acceptance. Those answer different questions, so do not switch between them silently.

Preserve definition versions. If qualification becomes stricter, show the break or restate comparable history where feasible. Otherwise, a process improvement may appear to be a demand collapse.

• Count each buying organisation once within the defined opportunity cycle.
• Exclude duplicates, spam, test records and existing opportunities.
• Record who accepts qualification and how disputes are resolved.
• Set a data-finalisation date and label provisional totals.

5. Separate the north star from its drivers and guardrails

One north star does not mean one number on the dashboard. Build a small driver tree showing the steps that produce the outcome. Drivers guide action; guardrails identify damage that a rising headline number could conceal.

For a hypothetical venue targeting attended midweek covers, drivers could include booking-page visits, booking completion and attendance. Guardrails might include contribution per cover, cancellations and service capacity. Discounting could lift attendance while undermining economics, so the headline alone cannot approve the offer.

Keep the tree small enough to support decisions. Assign an owner to each actionable driver, but do not imply the journey is perfectly linear. Walk-ins and telephone bookings need explicit treatment rather than being forced into the website funnel.

• North star: one shared outcome.
• Drivers: a few measurable steps the team can influence.
• Guardrails: quality, economics and capacity boundaries.
• Diagnostics: channel detail opened when something needs explaining.

6. Reconcile sources before combining channel results

Choose an authoritative source for the outcome, such as the order ledger for fulfilled orders. Advertising and analytics reports can explain acquisition, but adding their attributed conversions can double-count the same customer.

Google distinguishes GA4 key events from Google Ads conversions. Document the underlying event, counting rules and reporting context before comparing totals. A shared event name does not make reports interchangeable.

Use consistent campaign tagging to make acquisition analysis more interpretable. Google's manual campaign guidance explains how UTM parameters populate campaign dimensions. Tags describe traffic context; they do not prove that a campaign caused a sale.

In a hypothetical click-and-collect retailer, reconcile online purchases against fulfilled orders and cancellations. Report unlinked records separately. Do not allocate missing outcomes across channels merely to make a dashboard balance.

• Audit duplicate events, missing identifiers and reporting delays.
• Document attribution windows and source-specific definitions.
• Check consent-related gaps and offline outcome coverage.

7. Match the review cadence to the outcome's delay

A slow outcome needs faster diagnostic signals, not a different north star every week. For a hypothetical commercial fit-out firm, accepted projects may take months to emerge. Review qualified consultations weekly and project outcomes by enquiry cohort once enough time has elapsed.

Separate calendar reporting from cohort reporting. Projects signed this month may originate from much older campaigns. Comparing this month's spend with this month's signed value can misrepresent the acquisition process.

Set targets using baseline variation, available budget and delivery capacity. Treat them as planning assumptions, not guarantees. A before-and-after increase is an observation, not proof of marketing impact. Where suitable, Google Ads experiments can compare changes against an original campaign; that evidence remains scoped to the experiment.

• Weekly: inspect drivers, source health and guardrail breaches.
• Monthly: review the outcome, uncertainty and operational constraints.
• At an agreed strategy review: reconsider the definition, not just the target.

8. Measure efficiency without replacing effectiveness

A north star should reduce arguments about priorities, but that benefit needs measurement. Track reporting preparation time, decision latency and rework alongside the customer outcome. Fewer hours spent reporting is useful only if decisions remain reliable.

Hypothetically, if a weekly review falls from three hours to two, record one hour less for that review, not a company-wide productivity gain. Include extraction, reconciliation, approval and correction time. Compare similar workloads and note changes in staffing or scope.

Record where released capacity goes: testing a landing page, improving follow-up or handling more work. Time released is not automatically a cash saving. Labour costs only fall if expenditure actually changes, and faster execution does not guarantee better customer outcomes.

• Log hands-on time separately from waiting time.
• Measure correction rates and omitted checks.
• Avoid rewarding output volume when quality or outcomes deteriorate.

9. Turn the metric into a repeatable decision process

Pilot the definition before attaching incentives. Each review should identify the outcome, explain the most important driver change and select work with a plausible path to improvement. Record expected effects and guardrails before delivery so retrospective explanations cannot rewrite the original reasoning.

Its described workflow connects DailyRise recommendations, RiseProjects delivery and RiseInsights reporting. RiseInsights compares predicted and measured impact with sources and windows named; period comparisons do not establish causation.

See DailyRise, RiseInsights and the solution overview below. Connected evidence depends on permissions, provider access and source quality. TikTok and LinkedIn integrations remain approval pending, and Microsoft Advertising is planned. Do not build the measurement contract around unavailable data or assume unsupported business outcomes will appear automatically.

• Approve the definition, owner, source and review rhythm.
• Resolve quality problems before rewarding performance.
• Keep human judgement responsible for priorities and interpretation.