Skip to content

KPIs & Metrics

First PublishedLast UpdatedByAtif Alam

Upstream: Martech Stack & Automation — instrumentation, tooling, and data plumbing. This page covers measurement execution: which KPIs to pick, how to define them, and how to keep the data trustworthy.

Workbook cross-link: If you set KPI baselines in Demand Validation Experiments, use the KPI tree worksheet below to promote those priors into production metrics with owners, formulas, and review cadence.

KPIs (Key Performance Indicators) are the few numbers leadership actually watches — typically 5–12 at the company level, fewer at the team level.

Metrics are everything else you might look at when diagnosing why a KPI moved. Most metrics should never reach the board deck.

Use KPIs when you need alignment and accountability. Use metrics when you’re debugging a funnel step, a channel, or a cohort.

Every healthy measurement program has:

  1. One North Star Metric (NSM) — the single outcome that best captures customer value delivered.
  2. 3–5 input metrics — levers that predict movement in the NSM (signup rate, activation rate, etc.).
  3. Guardrail metrics — things that must not get worse while you optimize the NSM (churn, support tickets, gross margin).
flowchart TB
    NSM[North Star Metric]
    Input1[Input metric 1]
    Input2[Input metric 2]
    Input3[Input metric 3]
    Guard[Guardrail metrics]
    Diag[Diagnostic metrics]
    Vanity[Vanity metrics]

    Input1 --> NSM
    Input2 --> NSM
    Input3 --> NSM
    NSM --> Diag
    Diag --> Vanity
    Guard -.must not worsen.-> NSM

Vanity metrics (pageviews, raw followers, impressions without conversion context) belong at the bottom — explicitly labeled “don’t report these to the board.”

Vanity (activity)Actionable (outcome)Why it matters
ImpressionsMER (blended efficiency)Spend efficiency, not reach
Social followersSignup rate from socialAudience quality
PageviewsActivation rateTraffic that converts to value
Email opensTrial→paid from email cohortEngagement that drives revenue
MQL volumeMQL→SQL acceptance rateLead quality, not quantity

If removing the number wouldn’t change a weekly decision, it’s a metric — not a KPI.

  1. Pick one NSM — ask: “If this number grows sustainably, is the business healthy?” (Not: “What’s easiest to measure?”)
  2. Map 3–5 inputs — for each, name the owner, formula, data source, and target.
  3. Add 2–3 guardrails — e.g. gross margin, churn rate, NPS floor.
  4. Document every KPI on a metric-definition card (template below).
  5. Review monthly — kill KPIs nobody acts on; promote metrics that repeatedly drive decisions.
  6. Publish one page — the GTM Measurement Plan scorecard is the home for the live tree.
NORTH STAR METRIC
Name: _______________________
Formula: _______________________
Target: _______________________
Owner: _______________________
INPUT METRICS (3–5)
1. _______________ Formula: _______ Target: _______ Owner: _______
2. _______________ Formula: _______ Target: _______ Owner: _______
3. _______________ Formula: _______ Target: _______ Owner: _______
4. _______________ Formula: _______ Target: _______ Owner: _______
5. _______________ Formula: _______ Target: _______ Owner: _______
GUARDRAIL METRICS (2–3)
1. _______________ Floor/Ceiling: _______ Owner: _______
2. _______________ Floor/Ceiling: _______ Owner: _______
3. _______________ Floor/Ceiling: _______ Owner: _______
DIAGNOSTIC (not in board deck — for weekly ops)
- _______________
- _______________
Metric name: _______________________
Type: [ ] NSM [ ] Input [ ] Guardrail [ ] Diagnostic
Formula: _______________________
Data source: _______________________
Refresh cadence: _______________________
Owner: _______________________
Target: _______________________
Review cadence: _______________________
Last verified: _______________________
Notes: _______________________

Run quarterly. Score each 0 (missing) / 1 (partial) / 2 (complete).

Events & taxonomy (10)

  • Every funnel step has a named event
  • Event names follow snake_case convention
  • Properties documented in tracking plan
  • Signup event fires once per user (no duplicates)
  • Activation event matches product definition
  • Paid conversion event includes revenue amount
  • UTM parameters captured on landing
  • Cross-domain tracking configured
  • Mobile app events mirror web taxonomy
  • Server-side events for critical conversions

Identity (8)

  • Anonymous → known user merge works
  • CRM contact ID links to product user ID
  • Identity resolution rate measured
  • Duplicate account detection in place
  • Logged-out vs logged-in sessions distinguished
  • B2B account-level ID where applicable
  • Consent state stored with user record
  • Deleted-user handling documented

Data quality (10)

  • Null-rate monitored for key properties
  • Event volume anomaly alerts configured
  • Backfill process documented
  • Timezone consistent (UTC recommended)
  • Revenue currency normalized
  • Bot traffic filtered
  • Test/staging events excluded from production
  • Sampling documented (if used)
  • Data freshness SLA defined
  • Monthly reconciliation vs finance

Governance (12)

  • Tracking plan owner assigned
  • Change request process for new events
  • Deprecation process for old events
  • Dashboard owner per KPI
  • Metric definitions linked from dashboards
  • Access controls on PII
  • GDPR/CCPA deletion workflow
  • Consent banner coverage audited
  • Third-party tag inventory current
  • Martech stack diagram updated
  • On-call for tracking outages
  • Post-incident review template

Target: ≥70/80 before trusting KPIs for budget decisions.

Pick the pack that matches your motion and stage. These are starting sets — customize after 4–6 weeks of real data.

StageNorth Star candidateInput metricsGuardrails
Pre-PMFActivated teams/weekSignup rate, activation rate, time-to-first-valueSupport ticket volume
GrowthActivated teams/week or NRRMQL→SQL rate, trial→paid, CAC by channel, pipeline influencedChurn rate, sales cycle length
MaturityNRR or activated teamsCAC payback, expansion rate, win rate (influenced), MERGross margin, logo churn

See PLG vs Sales-Led Measurement if you run hybrid motion.

StageNorth Star candidateInput metricsGuardrails
Pre-PMFD7 retentionInstall→register rate, activation (first workout), CPI by channelCrash rate, app store rating
GrowthD7 or D30 retentionFree→paid %, MER, LTV/CAC, referral rateUninstall rate, refund rate
MaturityD30 retention or LTV/CACCohort NRR equivalent, expansion (premium tier), community engagementCPI inflation, ad fatigue signals

Caveat: Benchmarks are directional priors — not targets. Your category, price point, and motion change what’s “good.” Ship instrumentation, run 4–6 weeks, then set targets from your curves.

MetricPre-PMF rangeGrowth rangeNotes
Landing → signup2–8%5–15%Higher for narrow ICP landing pages
Signup → activation20–40%35–55%Activation definition matters enormously
Trial → paid5–15%10–25%Credit card at trial start adds 2–3×
MQL → SQL15–30%25–40%Depends on MQL scoring strictness
CAC payback (months)12–248–14Enterprise skews longer
NRR90–105%105–120%Below 100% = contraction problem
MetricPre-PMF rangeGrowth rangeNotes
Store listing → install15–35%25–45%ASO quality drives top of range
Install → register40–70%55–80%Friction on signup screen
Register → activation25–45%40–60%First workout completed
Free → paid (30-day)3–8%6–15%Annual plan mix helps
D7 retention15–30%25–45%Category-defining for fitness
D30 retention8–20%20–40%Community features lift this
MER (blended)0.3–0.80.8–1.5Below 1.0 = spending more than Year-1 revenue

Brand metrics belong in the KPI tree — usually as guardrails or diagnostic inputs, rarely as the North Star (unless you’re a pure brand play).

Brand metricTypical role in treeReview cadenceSource
Brand-NPSGuardrail (floor)QuarterlySurvey
Share of voiceDiagnostic inputMonthlySocial listening / SEO
Aided awarenessDiagnosticQuarterlyBrand tracking study
Unaided awarenessDiagnosticQuarterlyBrand tracking study
Consideration rateInput metricQuarterlySurvey + funnel

Integration rule: when performance KPIs improve but brand-NPS drops, pause scaling until you understand why. See Brand Perception for measurement methodology.

Bad data produces confident wrong decisions — worse than no data.

Minimum hygiene before reporting KPIs to leadership:

  1. Tracking plan exists — one doc listing every event, property, and trigger.
  2. Identity resolution works — you can connect anonymous visit → signup → paid in one user journey.
  3. Reconciliation monthly — product analytics signups ≈ CRM new contacts ≈ finance new customers (within agreed tolerance).
  4. Consent coverage — know what % of users accept analytics cookies; document impact on completeness.
  5. Anomaly alerts — event volume drops 30%+ → page someone.

Cross-link to Martech Stack for tooling layers and Funnel: Acquisition for UTM conventions that feed clean channel data.

Metrics to track (meta-metrics about your metrics)

Section titled “Metrics to track (meta-metrics about your metrics)”
Meta-metricTargetWhy
KPI definition coverage100% of board KPIs have documented formulasPrevents “which number is right?” debates
Data freshnessDashboards updated within 24h of eventStale data → stale decisions
Tracking coverage≥95% of funnel steps instrumentedCan’t optimize what you don’t measure
Reconciliation variance<5% between systems on signup countTrust threshold
Tracking-plan audit score≥70/80See audit template above

North Star: Teams reaching activation within 7 days (≥3 active users + 1 integration connected).

Input metrics:

  • Weekly signups (target: 420/week at growth stage)
  • Signup → activation rate (target: 38%)
  • Trial → paid (target: 12%)
  • MQL → SQL (target: 28% for sales-led segment)

Guardrails:

  • Logo churn <3% monthly
  • Support CSAT >4.2/5
  • Brand-NPS >35

Diagnostic (weekly ops, not board):

  • CAC by channel (Google $42, LinkedIn $180)
  • Landing-page CVR by campaign
  • Time-to-first-value median (target <2 days)

North Star: Users completing 3 workouts in first 14 days.

Input metrics:

  • CPI by channel (target: <$5 Meta, <$8 Apple Search Ads)
  • Install → register (target: 68%)
  • Register → activation (first workout) (target: 31%)
  • Free → paid within 30 days (target: 8%)

Guardrails:

  • D7 retention >28%
  • App store rating >4.5
  • Refund rate <2%

Diagnostic:

  • D1 / D7 / D30 retention curves by channel cohort
  • Onboarding quiz completion rate
  • MER blended (target: 1.1 at growth stage)
  • Too many KPIs. If everything is key, nothing is. Cap at 12 company-wide.
  • NSM that moves without revenue. “Daily active users” without monetization path is a vanity trap for paid products.
  • Undefined activation. Two teams measuring activation differently → permanent confusion. Write the definition once.
  • Benchmark worship. Industry averages ignore your price, motion, and ICP. Use benchmarks to sanity-check, not to set OKRs.
  • No owner. A KPI without a named owner is a metric nobody fixes.
  • Dashboard without cadence. See Reporting Cadence — data needs a rhythm.
  • Product analytics: Mixpanel, Amplitude, PostHog
  • Marketing analytics: GA4, HubSpot, Northbeam, Triple Whale
  • Metric catalogs: dbt metrics layer, Looker LookML
  • Books: Lean Analytics (Croll & Yoskovitz), Hacking Growth (Ellis & Brown)