📘 Budgeting: Paid Media

What you’ll learn

1. Unit Economics: The Gate on SpendDerive how contribution margin, CAC, LTV, LTV:CAC, payback period, and margin-based target CPA/ROAS set the ceiling on what you can afford to spend on paid media.Paid-media spend is gated by unit economics, not ambition. CAC is acquisition spend per new customer, while LTV should be built on contribution margin (revenue minus variable cost), since only margin can pay for advertising. The LTV:CAC ratio (a ~3:1 benchmark) and the CAC payback period together judge whether acquisition is both profitable and cash-efficient. From contribution margin and a required profit, you derive a target CPA and an implied target ROAS that bound every bid and budget.
2. Budget Mechanics: Daily, Lifetime, and PacingDistinguish daily versus lifetime budgets and explain pacing so you can match the budget mechanism to a campaign's intent while avoiding underdelivery and overspend.Daily budgets cap average spend per day and suit always-on campaigns, while lifetime budgets cap a fixed total across a defined flight and suit time-bound promotions. Pacing distributes spend over time and treats the daily budget as an average, so individual days can run above or below it. Underdelivery (spend stuck below budget) usually traces to restrictive bids, target CPAs, or small audiences, while overspend risk comes from front-loading and shared account budgets. Account caps and daily monitoring are the backstops that keep delivery efficient.
3. Diminishing Returns and Marginal ROASRead the spend-versus-efficiency response curve and use marginal ROAS, rather than average ROAS, to locate the profit-maximizing level of spend on a channel.As spend rises, total conversions usually keep growing but each added dollar buys fewer of them, because the most efficient demand is captured first and channels saturate. Average ROAS (total revenue over total spend) blends cheap and expensive conversions and hides this, whereas marginal ROAS measures the return on the next increment. Profit is maximized where marginal ROAS equals break-even ROAS; spending past that point grows revenue but erodes profit. Recognizing saturation explains why shifting budget to a fresher channel can beat piling more into a tired one.
4. Allocation Across Channels and the FunnelAllocate budget across channels and funnel stages (prospecting versus retargeting) by following marginal return while guarding against double-counted, self-reported credit.Budget should flow to wherever the next dollar earns the best marginal return, rebalanced on a regular cadence, not to whoever reported the best average last month. Across the funnel, prospecting creates demand at the top while retargeting captures it at the bottom, and funding only retargeting shrinks the future audience. Because platforms credit conversions independently, the sum of their reported revenue can exceed actual revenue, biasing budget toward aggressive credit-claimers. Reconciling platform reports against your own order ledger is essential before reallocating.
5. Incrementality Versus AttributionContrast last-click attribution with incrementality testing and use measured lift (iROAS) to redirect budget toward spend that genuinely causes sales.Attribution assigns credit for conversions, and last-click systematically over-rewards late touches like branded search and retargeting while ignoring demand creation upstream. Incrementality instead asks the causal question of what happened only because of the ad, best estimated with geo holdouts or randomized conversion-lift tests against a held-out group. Platforms over-credit themselves because they see only their own touchpoints and claim conversions within their windows, often sales that would have occurred anyway. Budgeting on incremental ROAS rather than reported ROAS shifts money toward channels that truly drive new revenue.
6. Scaling, the Learning Phase, and GuardrailsScale spend without resetting the learning phase, set guardrails and a reporting cadence, and use forecasting with scenario planning to manage the budget proactively.New or heavily edited campaigns enter a learning phase during which results are noisy; Meta notes ad sets typically exit after about 50 optimization events within roughly a week. Large sudden budget changes can re-trigger learning, so scaling in moderate, tested increments preserves earned optimization. Guardrails such as spend caps, ROAS floors, and automated pause rules contain risk when no one is watching, while a tiered cadence (daily anomaly checks, weekly trends, monthly strategy) ties each review to a decision. Forecasting with base, upside, and downside scenarios pre-decides responses to shifts in CAC, conversion rate, or marginal ROAS.
7. Case Study Critique: The Q4 Scaling PlanApply the course's frameworks to peer-critique a flawed Q4 paid-media plan, diagnosing its allocation, scaling, measurement, and guardrail weaknesses with quantitative reasoning.A peer team's $200,000, eight-week plan puts 80 percent into retargeting on a 6.0 last-click ROAS, doubles daily spend in week two, and reports only weekly. The critique shows the retargeting tilt confuses attribution with causation and starves prospecting, the overnight doubling risks resetting the learning phase and overshooting a ~1.67 break-even ROAS, and the weekly last-click reporting hides anomalies while ignoring incrementality. Stronger moves are funding prospecting, validating retargeting with a holdout test, scaling incrementally on marginal ROAS, and adding daily anomaly checks plus explicit guardrails. The exercise integrates unit economics, diminishing returns, allocation, incrementality, and scaling into one judgment.

Questions this course answers

Grounded in trusted sources

• IAB, Digital advertising overview, https://www.iab.com/
• Google Ads Help, Budgets and bidding, https://support.google.com/google-ads/
• Meta Business Help Center, Budgeting and bidding, https://www.facebook.com/business/help
• Harvard Business Review, Marketing ROI and spend allocation, https://hbr.org/

Every Wunder lesson is built from real, reputable sources — never invented.

Related courses