Paid Media KPIs - Toffu AI Skill

Apply this when reporting on paid media performance or choosing which metrics to surface.

Metrics That Drive Decisions

Marginal CPA - Not average CPA. The cost of the NEXT conversion at current spend levels. If your last 10% of spend generated 3% of conversions, you're in diminishing returns territory.
Blended CAC - Total marketing spend / new customers acquired, across ALL channels. The number your CFO cares about. Individual platform ROAS is an optimization metric, not a business metric.
Contribution Margin After Ad Spend - Revenue minus COGS minus ad spend. A product with 3x ROAS and 20% margin is break-even. A product with 2x ROAS and 60% margin is highly profitable. ROAS without margin context is meaningless.
New Customer vs. Returning - If 40% of your "conversions" are returning customers who would have bought anyway, your true acquisition ROAS is 40% worse than reported. Separate these in every report.

Platform-reported ROAS - Every platform over-attributes by 20-50%. Google counts view-through, Meta counts 1-day view, both take credit for organic conversions. Always cross-reference with backend revenue data.
CTR in isolation - High CTR with low conversion rate means your ad promises something the landing page doesn't deliver. Low CTR with high conversion rate is fine - the algorithm is finding converters, not clickers.
CPM - A market price, not a performance metric. CPMs rise in Q4, during elections, and any time auction demand increases. Judging a campaign by CPM is like judging a stock trader by the price of shares instead of returns.
Impressions and Reach - Proof of spending money, not proof of impact. Include in appendices for transparency, never in summaries.

Always compare to the same period last year, not just last month. Marketing has seasonality. Comparing January to December without context creates false crises or fake wins.
When ROAS improves but revenue is flat, spend decreased. When ROAS declines but revenue grew, you're scaling. Tell the full story - efficiency and volume together.
If a metric moved more than 30% week-over-week, check for data issues before reporting it as a real change. Tracking breaks, attribution windows shift, and weekday/weekend ratios vary.
Never average percentages across campaigns with different spend levels. A campaign with $100 spend and 10x ROAS averaged with a campaign with $50K spend and 1.5x ROAS does not yield 5.75x ROAS. Use spend-weighted calculations.