--- name: allocate-budget-across-campaigns description: >- Turns growth and efficiency goals into concrete daily budgets per campaign: computes the monthly spend floor from the goal math (target volume times target cost, or target revenue over target ROAS), stress-tests that number against real demand and historical efficiency, tiers campaigns by proven performance, and applies the resulting daily budgets. Use it at account/quarter kickoff, when goals change, or when budgets were set by feel rather than math. Not for rescuing a single campaign pinned by its cap (handle-budget-limited-campaigns), for pushing an already-allocated account past its current performance (scale-bids-and-budgets), or for setting CPA/ROAS targets (calculate-bid-targets). --- # Allocate Budget Across Campaigns ## Purpose Budgets should be derived, not guessed: the growth goal and the efficiency target jointly imply a minimum monthly spend, and historical performance tells you which campaigns have earned the largest share of it. This skill runs that derivation end to end, goal capture, spend-floor math, three feasibility checks, performance tiering, and the actual budget writes, so the account's money sits where the evidence says it works. ## When to run - New account, new campaign set, or client onboarding needing first budgets. - Quarterly/annual planning cycle, or goals changed materially. - Current budgets are historical accidents (never rebalanced, spread evenly, or set by request). - A weekly/monthly review shows systematic mispacing across the account. ## When NOT to run - One profitable campaign is hitting its daily cap -> handle-budget-limited-campaigns. - Budgets are sound and you want MORE volume/profit than the current stable state -> scale-bids-and-budgets. - The efficiency targets themselves are unvalidated -> calculate-bid-targets first (the spend-floor math divides by them; garbage in, garbage out). - You need pooled bidding or a shared budget mechanically set up -> set-up-portfolio-bid-strategies. ## Prerequisites - Documented growth goal (e.g. leads/month, revenue/month, new customers/month) and efficiency goal (target CPA / ROAS / acquisition-cost ceiling), confirmed with the stakeholder. - Validated unit economics (calculate-and-validate-unit-economics) so targets are above breakeven. - 3-6 months of account history where available; for cold starts, keyword-level forecasts stand in. ## Procedure 1. **Session guardrail.** Call `gads_policy_guardrail` before this session's first write. 2. **HUMAN STEP (outside VigilDog).** Confirm both goals with the stakeholder: one growth number, one efficiency number, each specific and measurable. Refuse to proceed with only one, growth without an efficiency guardrail burns margin; efficiency without a growth floor shrinks the account into irrelevance. 3. **Compute the monthly spend floor.** Pure arithmetic, no tools: - Volume-goal accounts: floor = target conversions x target CPA (e.g. 100 leads x 50 = 5,000/mo). - Revenue-goal accounts: floor = target revenue / target ROAS (e.g. 100,000 at 400% = 25,000/mo). - Subscription accounts: derive max acquisition cost from lifetime value (LTV / 3 is a common ceiling) or payback period (monthly revenue per user x acceptable payback months), then floor = target new customers x that ceiling. Add a working buffer: 15-20% when growth is the primary goal, 5-10% when efficiency is. 4. **Feasibility check 1, market.** Use `gads_keyword_ideas` and `gads_keyword_forecast` to confirm search demand can absorb the implied click volume, and `gads_run_gaql_query` to pull impression-share headroom (`metrics.search_budget_lost_impression_share`, `metrics.search_rank_lost_impression_share`) on existing campaigns. No headroom + no demand means the goal, not the budget, must change. 5. **Feasibility check 2, performance.** Pull 3-6 months of actuals via `gads_get_account_performance` and `gads_run_gaql_query`. If historical CPA/ROAS is far from the target used in step 3, the floor is fiction: either revise the efficiency target, fix upstream conversion rate, or reduce the growth goal. 6. **HUMAN STEP (outside VigilDog), feasibility check 3, business.** Stakeholder confirms: spend level approved, operations can absorb the projected volume, cash flow tolerates the payback window, and the Google Ads share of total marketing budget is deliberate. 7. **Tier the campaigns.** For every active campaign, pull trailing 30-90 day cost, conversions, CPA/ROAS, and IS-lost metrics via `gads_get_campaign` + `gads_run_gaql_query`, then classify: - Tier 1, proven: consistently at/inside target -> ~70% of budget. - Tier 2, scaling candidates: good efficiency plus visible headroom (IS lost to budget) -> ~20%. - Tier 3, tests: new types, new audiences, experiments -> ~10%. Shift the split toward 60/25/15 when growth is primary, 75/15/10 when efficiency is. 8. **Convert to daily budgets.** Daily total = monthly floor (with buffer) / 30.4. Distribute across campaigns per tier, and sanity-check each tCPA campaign's daily budget stays >= 10x its target CPA. 9. **Apply budgets (write).** Read current state with `gads_list_budgets`, then apply changes with `gads_update_budget` (existing) or `gads_create_budget` (new campaigns or a deliberate shared pool). Preview every change (validate_only=true default) as an old->new table across all campaigns; apply only after the user approves the whole allocation. Do not use shared budgets on campaigns that will run experiments. 10. **Set up monitoring.** Define the recurring pulls: daily spend vs budget (pacing %), IS lost to budget, IS lost to rank, all via `gads_run_gaql_query`. First two weeks: watch for budget-limited statuses and underspend. Week 3-4: compare tier assignments to actual results and promote/demote with fresh approvals. 11. **HUMAN STEP (outside VigilDog).** Warn the stakeholder that Google may spend up to 2x the daily budget on strong days (monthly totals still cap at ~30.4x daily); this prevents a week-one alarm. ## Decision rules - **The floor is a floor.** The computed monthly number is the minimum consistent with the goals, not a ceiling; overspending it is fine if efficiency holds. - **Buffer:** growth-primary 15-20%, efficiency-primary 5-10%. - **Tier split:** 70/20/10 baseline; 60/25/15 growth-tilted; 75/15/10 efficiency-tilted. Always keep at least ~10% in tests, an account with zero test budget has zero future Tier 1s. - **Daily conversion:** monthly / 30.4. - **Underspend diagnosis:** if a campaign cannot spend its allocation, the constraint is targeting or rank, not budget, broaden queries or fix competitiveness rather than shuffling money. - **Overspend-too-fast diagnosis:** high IS lost to budget on a profitable campaign justifies more budget; on an unprofitable one it justifies tighter targets, never more budget. - **Rebalance cadence:** monthly tier check (promote a Tier 2 outperforming Tier 1 economics), quarterly full recalculation, immediate recalculation when goals or unit economics change. - **No history?** Forecast clicks from `gads_keyword_forecast` CPC estimates, multiply by an expected conversion rate to project conversions, start conservative, recalibrate after 2-4 weeks of data. ## Common failure modes - **Budgets without goals.** Allocation by gut or by last year's numbers has no error signal. Steps 2-3 are non-optional. - **Even spreading.** Giving every campaign the same budget treats proven winners and untested ideas identically, the tier split exists precisely to break that symmetry. - **Skipping feasibility.** A mathematically correct floor the market cannot absorb produces chronic underspend and a credibility problem; check demand before promising volume. - **Set-and-forget allocation.** Tier membership drifts; a quarterly recalc plus monthly tier review keeps money following evidence. - **Shared budget + experiment collision.** Campaign experiments require individual budgets; putting an experiment campaign on a shared pool silently breaks the test design. ## Related skills - Before: calculate-bid-targets (efficiency targets), calculate-and-validate-unit-economics (breakeven inputs), select-a-bidding-strategy (strategy per campaign). - After: handle-budget-limited-campaigns (single-campaign cap issues), scale-bids-and-budgets (growing past the allocation), monitor-and-maintain-bid-strategy-health (recurring checks), set-up-portfolio-bid-strategies (if a shared budget + pooled strategy makes sense).