Budget Allocation Rules - Toffu AI Skill

Apply these rules when recommending budget changes or planning ad spend allocation.

Core Principles

Every dollar should be allocated based on marginal efficiency, not average performance. A campaign with $10K spend and 4x ROAS may drop to 2x ROAS if you push it to $20K. The average looks great; the marginal return is mediocre.
Budget allocation is a portfolio decision. Treat platforms like assets - diversify for resilience but concentrate for growth. A single-platform strategy is fragile; a five-platform strategy is unfocused.
The goal is maximum profitable spend, not maximum ROAS. A campaign at 2x ROAS spending $50K generates more profit than a campaign at 8x ROAS spending $2K. Chase total margin, not efficiency ratios.

Before increasing budget, check if the campaign is impression-share constrained (Google/Bing) or if frequency is still low (Meta/LinkedIn). If neither, more budget just buys worse traffic.
Scale spend by duplicating what works into new audiences or geos, not by increasing budget on a saturated campaign. Horizontal scaling beats vertical scaling past a certain point.
The 20% weekly budget increase rule exists because Meta and Google algorithms react badly to sudden changes. But if you're launching a new campaign with proven creative, start at the target budget immediately. The rule applies to optimization, not launch.
Seasonal demand changes are not optional to plan for. If you sell B2B, Q4 budgets should shift from acquisition to pipeline acceleration because decision-makers disappear in December. If you sell DTC, Q4 is when you go heavy.

Don't kill underperforming campaigns the same day they dip. Algorithms have volatility. Look at 7-day rolling average before cutting. But don't wait 30 days either - that's just burning money to avoid a decision.
When cutting, reduce budget rather than pausing. Pausing resets the algorithm's learning. A 50% budget cut preserves optimization data; pausing destroys it.
If a platform's blended ROAS is below 1x for 14 consecutive days with no structural change (new landing page, new creative, new tracking), it's time to reallocate. That's not a dip, that's the new baseline.

Reserve 10-15% of total budget for testing, but define "test" precisely. A test has a hypothesis, a metric, a timeline, and a kill criteria. "Let's try TikTok" is not a test. "We'll spend $3K on TikTok over 3 weeks targeting X, measuring CPA against our Meta benchmark of $Y, and kill it if CPA exceeds 2x" is a test.
New platforms need 2-3 months and $5K-15K minimum to generate statistically meaningful data. If you can't commit that, don't start.