If finance can't defend it, it isn't a metric

Operators get sold dashboards constantly. Vanity metrics, “engagement” scores, brand-health indices, NPS, attribution-light “MQL pipeline.” Most of these don’t survive contact with finance. The CFO asks one question — “where does this number come from?” — and the marketing team answers in three different ways.

We have a rule: if finance can’t defend the metric to the board, we don’t deliver projects against it. Not “we won’t measure it” — we just don’t anchor a price or an outcome to a number that can’t survive an honest financial audit.

What “finance can defend” means

Finance can defend a number when three things are true.

One — The definition is documented

There’s a written definition of the metric. Not a tooltip in a dashboard, an actual document. “Active matter” means a matter where the most recent stage transition occurred within the last 30 days. “Attributable revenue” means revenue from orders where the conversion event fired server-side and the timestamp is within the attribution window. The definition lives somewhere stable so it can’t drift.

Two — The data sources are reconciled

The metric flows from a known set of sources, and those sources reconcile to a system of record. If the metric is “monthly recurring revenue”, the source is the billing system, and the billing system reconciles to the bank weekly. If the metric is “lifetime customer value”, the calculation is run from the warehouse, the warehouse pulls from the billing system, and the billing system reconciles to the bank weekly. Every step has a reconciliation point.

Three — The drift is bounded

Week-over-week drift has a known band. “Cart-to-purchase rate normally moves ±2.5% week over week” is a defensible statement. When the number steps outside the band, finance investigates. When it stays inside, finance trusts the trend. The band itself is documented and reviewed quarterly.

Metrics that pass the test

The list is shorter than most operators expect.

Attributable revenue (server-side tracked, reconciled to bank). Cart-to-purchase rate (cohort, not averaged). Lead-to-close (with the lead-source attribution model documented). Hosted uptime against an SLA the operator can quote in proposals. Mean time to resolution on incident tickets, with incident definition explicit. Cost per acquisition with full attribution model. Average order value, weighted by cohort.

Each of these has a finance-grade definition the CFO already trusts or can be taught to trust in one meeting.

Metrics that don’t pass

Most “engagement” metrics. Most “brand health” metrics. Most NPS unless it’s run rigorously with sampling discipline. “Pipeline velocity” without an explicit stage-transition model. “Marketing-attributed revenue” when the attribution model isn’t documented or doesn’t reconcile to finance.

These metrics aren’t useless — they often correlate with the defensible ones. But they shouldn’t be the metric a project is priced against, because they can’t survive a CFO question.

What to do with non-defensible metrics

Archive, don’t delete. The vanity metric stays on a secondary tab somewhere. We instrument it (cheap), we report it (cheap), but we don’t price the project against it. When the project closes, we report the defensible metric in the closeout document; the vanity metric is a footnote.

Operators sometimes push back here. “We track NPS, we want to see NPS in the report.” Fine — NPS goes in. We just don’t make the project contingent on NPS moving, because we can’t defend the cause-and-effect chain to finance.

Why this matters commercially

Two reasons.

One — It changes which projects close

When an operator scopes against a defensible metric, the CFO can sign the engagement letter without a debate. When the metric is vanity, the project gets stuck in legal-and-finance review for weeks, often dies there. We closed a higher percentage of scoped projects once we adopted this rule, even though we walked away from more.

Two — It changes which projects get repeat business

The closeout meeting is the next-project-or-not moment. If the metric we hit was defensible, finance is satisfied, the operator is satisfied, and the next project is easy to scope. If the metric was vanity, the closeout meeting has a “did this actually work?” subtext that can’t be resolved. We see fewer second projects from those.

The implementation

It’s a one-page rule, not a methodology. Before pricing any project, we write the metric definition. We trace the data sources. We document the drift band. We make sure the CFO would sign off on the scope.

If we can’t get to a defensible metric, we re-scope until we can. Sometimes that means narrowing — instead of “lift conversion”, we scope “lift cart-to-purchase from 32% to 38% on the mobile cohort.” Sometimes that means broadening — instead of “improve NPS”, we scope “reduce mean time to resolution on support tickets from 6 hours to 2.” Either way, the metric ends up defensible.

The takeaway

Finance is the gate. Every project metric clears it before pricing. The discipline costs us some projects at scoping, returns more projects at closeout, and produces a track record we can publish as case studies. Net win.