Budget Shifter

1. Purpose

When a campaign wins, raising its budget in isolation is half the job — the other half is finding the money without blowing total account spend. This skill identifies donors (underperformers that can shed budget without hurting delivery) and receivers (winners with room to scale), then proposes a set of moves with amounts and reasoning.

Two modes:

• Redistribute (default): total account spend unchanged; money moves between campaigns.
• Scale up: additive — suggest incremental budget for receivers without donor cuts.

2. Donor Signals (pull budget from)

A campaign is a donor if it hits any of these:

| Signal                    | Threshold                                                          |
| :------------------------ | :----------------------------------------------------------------- |
| **Underperforming ROAS**  | ROAS < 0.8× account median (last 14 days)                          |
| **Overshooting CPA**      | CPA > 1.3× target CPA for 7+ days                                  |
| **Fatigued at scale**     | Frequency > 3.5 AND spending > 10% of account daily total          |
| **Underdelivering**       | Spending < 70% of its budget daily for 5+ days — budget is unused  |

3. Receiver Signals (send budget to)

A campaign is a receiver only if it meets all of these:

| Signal                  | Threshold                                                                          |
| :---------------------- | :--------------------------------------------------------------------------------- |
| **ROAS above target**   | ROAS > 1.2× target for 7+ days                                                     |
| **Delivery headroom**   | Daily spend at ≥ 85% of current budget — the campaign is actually using its money  |
| **Audience freshness**  | Frequency < 2.5                                                                    |
| **Past learning phase** | ≥ 50 conversions in trailing 7 days OR ad set out of platform "Learning" status    |

4. Move Sizing & Guardrails

| Rule                       | Value                                                                                    |
| :------------------------- | :--------------------------------------------------------------------------------------- |
| **Max single move**        | 20% of the receiver's current daily budget per shift                                     |
| **Donor floor**            | Never cut a donor below the learning threshold (~50 conversions/week/ad set)             |
| **Cooldown**               | Don't shift the same (source, destination) pair twice in < 48 hours                      |
| **Aggregate cap per run**  | Total reallocation ≤ 10% of account daily spend                                          |
| **New‑campaign shield**    | Campaigns < 7 days old are neither donors nor receivers                                  |
| **One‑way donor rule**     | A donor in one run can't become a receiver in the next 48h (avoid churn)                 |

5. Operational Output: Budget Shift Proposal

Budget Shifts — Mon, Apr 22 · Mode: Redistribute · Total moved: $180/day (4.5% of $4K account daily spend)

| From → To | Amount | Why | | :------------------------------------------------------ | :------- | :------------------------------------------------------------------------------------------------------------------------------------ | | Q1 Test Campaign → Retargeting — Conversions | $40/day | Donor: CPA $38 vs. $25 target (7 days). Receiver: ROAS 3.4× (target 2.0×), freq 1.8, past learning. | | Evergreen Prospecting → Founder UGC Hook | $60/day | Donor: underdelivering (52% of $120 budget used last 5 days). Receiver: 17 conversions in 7 days, CPA $15. | | Spring Sale Carousel → Summer Launch — Traffic | $80/day | Donor: fatigued (freq 4.1, CTR 0.8%, spending 14% of account). Receiver: CPC 42% of account avg, freq 2.1, spending 93% of budget. |

Action: apply all 3 in one batch, then re‑check in 48h before the next shift.

6. Scenario Forecasts

Beyond individual shift proposals, the skill answers what‑if questions grounded in the account's actual history — never generic projections. Every number in a forecast must be derivable from this account's data.

6.1 Supported Scenarios

| Scenario                       | Typical Question                                                          |
| :----------------------------- | :------------------------------------------------------------------------ |
| **Total spend ±X%**            | *"What happens if we spend 30% more next week?"*                          |
| **Single‑campaign scale**      | *"What if we doubled Retargeting's budget?"*                              |
| **New campaign launch**        | *"If we add a $200/day campaign, where does it bite into existing performance?"* |
| **Campaign pause**             | *"If we pause Summer Launch, where does its spend need to go?"*           |
| **Target‑driven**              | *"What spend gets us to $50K revenue/week at ≥ 2.0× ROAS?"*               |

6.2 Data Required Before Forecasting

Pull these from the account — if any are missing for a given campaign, skip or mark that campaign as 🔴 Unreliable in the output:

• Per‑campaign spend → outcome curve (14‑day rolling): observed ROAS and CPA at each realized daily spend tier.
• Past scaling response: actual % change in ROAS/CPM when that campaign's daily budget moved ±20% in the trailing 90 days.
• Saturation state from audience-health: does the campaign have headroom, or is it already Exhausted?
• Account auction context: current CPM vs. 30‑day avg (rising CPM ⇒ less elasticity at higher spend).

6.3 Forecast Rules

• Per‑campaign elasticity — never blend across campaigns. A Winner with headroom and a saturated brand campaign respond very differently to the same % increase.
• Bound to history — cap any projection at the historical extreme observed. Don't extrapolate past what this account has actually done.
• Confidence tiers:

| Tier                    | Criteria                                                                                   |
| :---------------------- | :----------------------------------------------------------------------------------------- |
| 🟢 **High**             | ≥ 3 past budget changes to sample, similar CPM regime, current saturation Healthy/Refreshing |
| 🟡 **Low**              | 1–2 past samples, or CPM regime differs > 20% from sample window                            |
| 🔴 **Unreliable**       | No relevant history (new campaign, no prior budget changes) — show range only, no point estimate |

6.4 Operational Output: "What if we spend 30% more?"

Scenario: account daily spend $4,000 → $5,200 (+30%) · horizon: next 7 days

| Campaign | Current | Proposed | Projected ROAS | Projected CPA | Conf. | Notes | | :-------------------------------- | :------- | :--------- | :------------------- | :-------------------- | :---- | :-------------------------------------------------------------------- | | Retargeting — Conversions | $1,400 | $1,820 | 3.1× (was 3.4×, −0.3)| $19 (was $18.50, +$0.50) | 🟢 | Last 2 +20% moves held ROAS within −5%. Freq 1.8 — still headroom. | | Summer Launch — Traffic | $1,050 | $1,260 | — (traffic objective)| CTR flat | 🟢 | CPM only +6% on last +25% move. Freq 2.1. | | Evergreen Prospecting | $320 | $700 | 1.4× (was 2.0×, −30%)| $42 (was $30, +$12) | 🟡 | Only 2 prior budget changes. Likely hits saturation at this tier. | | Founder UGC Hook | $280 | $420 | 2.8× (was 2.9×) | $15 (was $15.20) | 🟢 | Winner: 17 conv / 7d, freq 1.8, past learning phase. |

Aggregate forecast: account ROAS 2.4× → 2.2× (−8%) · CPA $24 → $26 (+8%) · +21 conversions/day.

Caveat: Evergreen Prospecting's elasticity sample is thin (🟡). Consider redirecting its +$380 to Retargeting + Founder UGC Hook (both 🟢) — projected aggregate ROAS with that redirect: 2.3× (vs. 2.2× baseline).