I teach a version of this topic in Growth Intelligence because teams rarely suffer from a lack of metrics. They suffer from too many metrics with unclear decision rights.
A KPI earns its place when it can change an action. If nobody knows what decision follows when a metric moves, it is probably reporting context rather than a key performance indicator.
Start with the decision, then choose the metric
“Improve marketing efficiency” is too vague. A real operating question sounds like: should we move another $10,000 from Meta to Google? Should we keep funding this creator program? Should we raise budget while CAC is increasing? Should sales respond within five minutes or thirty?
Once the decision is explicit, the KPI hierarchy becomes easier to design.
Three layers of KPIs
1. Business outcomes
Revenue, contribution margin, new customers, qualified pipeline, retention and Payback Period. These are the metrics leadership ultimately funds.
2. Funnel economics
CAC, cost per qualified lead, MQL-to-SQL rate, Win Rate, average order value, repeat rate and conversion rate. These explain how the business outcome is being produced.
3. Diagnostic signals
CPM, CTR, CPC, hook rate, video retention, landing-page engagement and frequency. These are valuable because they help diagnose the mechanism. They should rarely become the executive objective by themselves.
The KPI must match the funnel stage
Awareness activity should not be forced into a last-click conversion lens, but awareness also should not become immune to accountability. Use reach quality, attention, search lift, direct traffic, consideration studies and later-stage assisted effects.
Conversion campaigns should move closer to business truth: new-customer revenue, contribution margin, qualified leads and closed-won value.
Good KPIs have thresholds
A dashboard without thresholds produces observation. An operating dashboard needs decision rules. For example:
- If Blended CAC is more than 15% above target for two consecutive weeks, reduce acquisition spend or repair conversion.
- If MQL-to-SQL falls below the historical range, audit targeting and sales response before increasing media.
- If frequency rises while conversion rate declines, refresh creative before expanding budget.
- If platform ROAS rises but ecommerce revenue does not, investigate attribution overlap.
One metric should have one owner
Metrics become political when ownership is vague. Marketing may own lead generation, sales owns response and closing, product owns checkout friction, finance owns margin definitions. The executive system needs the handoffs documented so a miss does not become a debate about whose dashboard is correct.
The final question
At the end of a weekly review, ask: what did the data make us do differently? If the answer is “nothing,” the reporting layer is consuming time without improving the system.