| name | kpi-definition |
| description | Define KPIs that survive cross-team disagreement — single owner, single formula, single source of truth, decay-tested, documented in a KPI dictionary. Reach for this skill when launching a new metric, when two teams cite different numbers for "the same" KPI, when an exec asks "why doesn't this tie?", or when standing up a KPI pack. Used by `fpa-analyst` (primary) and `board-pack-composer`. |
Skill: kpi-definition
Purpose: Operational KPIs decide hiring, comp, board confidence, and capital allocation. A KPI without a defensible definition is a fight waiting to happen. Used by fpa-analyst (primary).
When to use
- Standing up a new KPI for a business / function
- Resolving "ARR is X" vs "ARR is Y" between two teams
- Standing up a KPI pack or operating cadence
- Pre-audit when a non-GAAP measure is reported externally
- Pre-board / pre-investor cycle to lock the period's definitions
- Annual / quarterly KPI dictionary refresh
The KPI dictionary entry
Every KPI ships with a dictionary entry. Anything less is folklore.
# KPI: <name>
**Definition:** <one-sentence plain English>
**Formula:** <unambiguous arithmetic / SQL>
**Source data:** <system + table + filter + as-of>
**Owner:** <named person + role>
**Refresh cadence:** <daily / weekly / monthly / quarterly>
**Inclusion / exclusion rules:** <what's in, what's out, with reason>
**Effective date / version history:** <YYYY-MM-DD definition v1; YYYY-MM-DD revised to v2 because Z>
**Cross-checks:** <what reconciles to this metric and at what frequency>
**Known limitations:** <what this KPI does NOT tell you>
Six properties of a defensible KPI
A KPI that passes scrutiny has all six:
| Property | Test |
|---|
| Unambiguously defined | Two analysts with the same source data compute the same number |
| Strategically relevant | Ties to a documented company strategy or operating objective — not a vanity metric |
| Sensitive to action | Leadership can move this in 1-2 quarters via decisions they control |
| Trend-readable | A multi-period view tells a story, not just noise |
| Reconciled | Reconciles or cross-checks against at least one other authoritative source (GL, billing system, CRM) |
| Auditable | Source query / extract / formula is reproducible without consulting the original author |
KPIs that fail two or more of these are candidates for retirement.
Canonical KPI definitions: the SaaS set
These are the most-contested KPIs in finance. Each comes with a sharp definition + the common mistake.
ARR — Annual Recurring Revenue
Definition: Annualized contracted revenue from recurring subscriptions at the measurement date.
Formula:
ARR = Σ (active subscription's monthly recurring revenue × 12) at measurement date
Include: committed subscription revenue under signed contracts with active service.
Exclude: one-time services, set-up fees, professional services, hardware sales, expired contracts not yet renewed, contracts in opt-out / cancellation.
Common mistake: counting deferred revenue release as ARR (that's a P&L item). ARR is a snapshot; deferred revenue is a roll. Different things.
Cross-check: ARR end-of-period should reconcile to billing-system + CRM contract data; not to GL revenue (which has timing and recognition differences).
NRR / NDR — Net Revenue Retention / Net Dollar Retention
Definition: Trailing-12-month ARR retained + expanded from a fixed customer cohort, divided by the cohort's starting ARR.
Formula:
NRR_t = (Cohort_ARR_t) / (Cohort_ARR_(t-12))
Where:
Cohort = customers in the book on the measurement date 12 months prior
Cohort_ARR_t = current ARR from those same customers (downgrades + churn + expansion all in)
Common mistake: including new customers acquired during the period. NRR is cohort-locked.
Industry benchmarks (2026): > 110% is good, > 120% is excellent, > 130% is unusual.
Gross Logo Retention
Definition: Percent of customers in the starting cohort still active at end of period.
Formula:
Gross Logo Retention_t = Cohort active customers_t / Cohort customers_(t-12)
Common mistake: weighting by ARR (that's revenue retention, not logo retention).
LTV, CAC, LTV/CAC, Payback
Customer LTV (subscription):
LTV = Gross profit per customer per month × (1 / monthly gross logo churn rate)
CAC:
CAC = (Sales + Marketing expense in period) / (New logos acquired in same period)
Payback period:
Payback (months) = CAC / Monthly gross profit per customer
Common mistake: including expansion in CAC denominator (expansion isn't customer acquisition). And: claiming LTV/CAC > 5 in early-stage businesses where the LTV horizon assumes 10-year retention before 3-year data exists.
Magic Number
Magic Number = (Net new ARR × 4) / S&M expense in prior quarter
A measure of sales efficiency. > 1.0 generally signals investing-mode; < 0.5 signals sales motion stress.
Cash Burn / Net Burn
Net burn = Beginning cash - Ending cash (period)
(excluding non-operating events like fundraises, M&A consideration)
Common mistake: including a fundraise inflow in net burn. Separate operating burn from financing flows.
Operating KPIs by department
Sales
- Quota attainment — sum of attained / sum of assigned, by rep, by team, by period
- Pipeline coverage — open pipeline value / quarterly quota target (typical guidance: 3-4× coverage)
- Win rate — deals won / deals decisioned (won + lost), with conversion-stage segmentation
- Sales cycle — median days from Stage X to Closed-Won, with quartile distribution
Marketing
- MQL → SQL conversion rate — qualified leads passed to sales / total marketing leads
- CPL by channel — cost per qualified lead, segmented by acquisition channel
- Marketing-sourced pipeline % — pipeline $ attributed to marketing-originated leads / total pipeline $
- Brand vs demand split — top-of-funnel awareness investment / total marketing investment
Customer Success
- Onboarding TTV — median days from contract signature to first value-event (defined per product)
- CSAT / NPS — survey scores with response-rate context (sub-30% response rates are noisy)
- Renewal rate — logos renewed / logos eligible for renewal in the period
Engineering / Product
- Velocity — story points or items shipped per sprint, trended (avoid quarter-to-quarter comp due to org changes)
- Cycle time — median time from work-started to production-deployed
- Reliability — error rate, uptime %, mean-time-to-recover
- Adoption — DAU / MAU, feature-adoption % among eligible users
Operations / Finance
- Cash conversion cycle — DSO + DIO − DPO (lower = better cash discipline)
- Operating margin — operating income / revenue
- Rule of 40 (SaaS) — Revenue growth % + Operating margin %; > 40 is industry-good
Reconciliation discipline
Every KPI in the dictionary lists at least one cross-check:
| KPI | Reconciles to |
|---|
| ARR | Billing-system contract data + CRM closed-won data |
| GAAP revenue | GL revenue accounts |
| Deferred revenue | GL balance + billing-system unrecognized billings |
| Customer count | CRM active accounts |
| Headcount | HR system payroll feed |
| Cash | Bank statements + GL cash |
When two systems disagree, the resolution is documented (not the disagreement papered-over). State which is authoritative for the KPI and why.
Versioning + change discipline
KPI definitions can change. When they do:
- Bump the version (e.g., ARR v1 → ARR v2)
- Document the change, the date, and the reason
- Restate prior periods under the new definition for trend comparability (or footnote the discontinuity prominently)
- Communicate to dependent reports (board pack, investor update, KPI pack)
Anti-pattern: changing a KPI definition silently mid-year. Auditors and boards lose trust faster from definition changes than from bad numbers.
Common failure modes
- One KPI, three definitions — Sales says ARR is X (includes services), Finance says Y (excludes services), Board pack uses Z. Define once; all reports inherit.
- Vanity metrics — "Total registered users" with no path to revenue. Strategically irrelevant.
- Insensitive metrics — leadership cannot move it in a quarter (e.g., "brand awareness in untested markets"). Useful for tracking, not for OKRs.
- KPI without an owner — nobody refreshes it; over time it drifts from reality.
- No effective date — when did this definition start applying? Auditors will ask.
- Missing limitations — every KPI has limitations. CSAT in a 5% response sample tells you less than the response-weighted average over 5 quarters.
- Cross-check absent — the KPI floats; one bad query and the number is wrong forever.
KPI hygiene checklist
Before a KPI enters the dictionary:
See also