The average ad account tracks forty metrics; the average budget decision uses three. The gap between measurement and decision-making is where performance leaks. This is the KPI hierarchy we install — a short list where every metric has an owner, a threshold, and an action attached.
The hierarchy: four levels, twelve metrics
| Level | KPIs | Governs |
|---|---|---|
| Business | Revenue, MER (total revenue ÷ total ad spend), blended CPA | Whether marketing works at all |
| Funnel | Qualified-lead rate, lead→sale rate, AOV/LTV by source | Where the funnel leaks |
| Channel | CPA/CAC by channel, marginal CPA, share of new customers | Budget allocation |
| Tactical | CTR, CPM, frequency, hook rate, landing conversion rate | Creative and placement fixes |
The rule: decisions flow up the table, diagnostics flow down. A tactical metric (falling CTR) earns budget action only through its channel-level effect (rising marginal CPA) — never directly.
Metric definitions that prevent arguments
- MER over ROAS: return on ad spend fragments by platform; marketing efficiency ratio sees the whole board and cannot be gamed by last-click windows.
- New-customer share: a channel "performing" at 4x ROAS while reselling existing customers is a retention expense wearing acquisition clothing.
- Marginal, not average CPA: what the next unit of budget buys — the only number that should move allocation.
- Payback period for subscription/LTV models: cash-flow reality that ROAS ignores.
Dashboard discipline
- One weekly scorecard: business level + channel level, same template every week — comparability beats comprehensiveness.
- Thresholds with assigned actions before the data arrives: "frequency > 2.5 → refresh hooks," not ad-hoc reactions.
- Annotated timeline: every structural change (tracking migrations, big creative shifts, budget moves) marked on the chart — memory is the cheapest analytics tool.
- Monthly: retire one metric nobody acted on in 90 days. Dashboards bloat silently.
The anti-patterns
- Reporting platform-reported numbers to leadership without calibration factors (see reconciliation).
- Vanity north stars: impressions, followers, engagement rate — diagnostics at best, never goals.
- Real-time dashboards as theater: weekly decisions need weekly granularity; hourly granularity mostly manufactures anxiety.
- KPI inflation: adding metrics to look rigorous — the surest sign a program has no governing numbers.
ROAS or MER?
Both, with jobs: ROAS optimizes within platforms; MER governs the whole program. When they disagree, MER is right.
How many KPIs belong on an executive dashboard?
Five to seven. Business level plus one line per channel. Anything more is a working dashboard wearing an executive costume.
Decide what runs the program — or the loudest metric will. Build the scorecard with us.