ads-setup-engine

Ads Setup Engine

Builds a custom Meta ABO testing campaign structure for lead gen accounts. The strategy is fixed — ABO testing with 3 hook variations per ad set, manual budget band management, kill/scale based on bands. What's NOT fixed is the specific dollar amounts of each band — those are calculated based on the user's account.

This skill is lead gen only. Direct purchase / e-com accounts use a different structure (CBO or Advantage+). Flag this upfront and refuse to run for e-com.

---

The Strategy (Non-Negotiable)

1. One campaign type: ABO Testing — ad set budget optimization, not CBO
2. One ad set per concept — each concept from the creative strategy becomes its own ad set
3. 3 ads per ad set — the 3 hook variations from the Script Generator are launched together inside the ad set
4. Starting budget: $50/day per ad set (standard — some start at $10/day, this system jumps straight to $50)
5. Budget bands govern kill/scale — not gut feel, not CBO delegation to Meta
6. One band change per ad set per day — tracked via "date last edited" column
7. Weekly + Daily cadence — weekly for long-lookback decisions, daily 3-day rolling for reactive adjustments

---

What You Need Before Starting

1. context/ad-account.md filled in (from Guided Setup Phase 3 + KPI Tracker lock-in)
2. At least one creative strategy brief at outputs/creative-strategy-*.md
3. User has access to their Meta Business Manager (for implementation)

If either of the first two is missing, stop and tell the user to complete foundations setup first.

---

Phase 1: Funnel Gate (1 min)

First question, before anything else:

"Before we build your campaign structure — what are you running? Lead generation (people book a call / fill a form / apply) or direct purchase (people buy a product on the landing page)?"

If they say direct purchase / e-com / shop / Shopify / checkout:

"This skill is for lead gen accounts only. The strategy we use — manual ABO testing with budget bands — works best when you have slow feedback loops and expensive conversions. For direct purchase, Meta's Advantage+ or broad CBO with the pixel firing at checkout usually beats manual management. I can't build this for you. Pause here and run a different campaign strategy."

Stop. Do not proceed.

If they say lead gen, continue to Phase 2.

---

Phase 2: Account Interview (10 min)

Ask ONE question at a time. Push back on vague answers.

Q1 — Target Monthly Spend

"What's your target monthly Meta ad spend over the next 90 days? Give me a specific number."

Convert to daily: monthly ÷ 30. Use this as the working "target daily budget" for all band calculations.

Q2 — Current CPR (Cost Per Result)

"What's your current or most recent cost per result? A result means a qualified booked call for lead gen. Rough number is fine."

If they don't know, pull from context/ad-account.md. If still unknown, push back:

"I can't build the bands without this. Run your last 30 days of data through the KPI Tracker skill first to establish this baseline. Come back when you know."

Q3 — Target CPR

"What CPR are you aiming for? This should come from your KPI Tracker output — it's the maximum allowable cost per result that keeps you profitable."

Pull from context/ad-account.md if they don't remember. If target is lower than current, note the gap — they need to close it through creative, not budget tricks.

Q4 — Account Average CPR

"What's your account-wide average CPR? Not your best ads, not your worst — the average across everything that spent money in the last 30 days."

This becomes the centreline of the bands. Most ads sit here; winners are below, losers above.

Q5 — Concept Volume

"How many concepts are you launching right now from your creative strategy brief? (A concept = one persona + angle + format combination.) Each concept becomes one ad set."

Pull from outputs/creative-strategy-*.md if uncertain — count the rows in the strategy map.

Q6 — Creative Production Cadence

"How many new concepts can you realistically produce per week going forward? Scripts, recorded, edited, uploaded. Be honest."

This determines how aggressive the kill rules can be. If they can only produce 2/week, we can't kill 10 ads a week or the account goes dark.

Q7 — Risk Tolerance

"On a scale of 1-5: how aggressive do you want the scaling rules to be?

• 1 = Conservative (small budget increases, longer windows before acting)
• 3 = Balanced (default — recommended for most)
• 5 = Aggressive (fast increases, fast kills — only if creative pipeline is deep)"

Default to 3 if they're unsure.

Q8 — Optimization Event

"What conversion event are you optimizing the pixel on? Book call / application submitted / lead form complete / other?"

If they don't know, tell them:

"Set this up before launching. The optimization event is what Meta chases — wrong event = wasted spend. Use the event closest to a qualified lead (usually 'Lead' or 'Schedule' or a custom 'Qualified Booking')."

---

Phase 3: Band Calculation (5 min processing)

Now calculate the 8 bands.

The Logic

Reference bands (for $35K/day target spend):

• New ad with 0 results: $10/day
• 1 result (CPR > $850): $10
• 1 result (CPR $700-850): $50
• 1 result (CPR $500-700): $100
• 2+ results (CPR > account avg): $500
• 2+ results (CPR < account avg): $1,000
• Top performers: $2,000
• Max: $3,000

standard modifications:

• Starting band = $50/day (not $10) — matches our concept = 3 hooks launched together
• Second result band scales from $50 to $100 (not $10 to $50)
• All subsequent bands scale proportionally to the user's target daily budget

The Calculation

Compute based on user inputs:

Target daily budget = monthly_spend / 30 (from Q1) Account average CPR = user's answer (from Q4) CPR tiers:

• Tier A = Top performers: CPR ≤ 60% of account average
• Tier B = Above average: CPR 60-100% of account average
• Tier C = At average: CPR 100-140% of account average
• Tier D = Below average: CPR 140-170% of account average
• Tier E = Fail: CPR > 170% of account average OR ≥ 2× target CPR

Apply risk multiplier (from Q7):

• Risk 1: multiply top-tier bands by 0.5x
• Risk 3: 1.0x (baseline)
• Risk 5: 1.5x

Calculate the 8 bands:

#	Condition	Band Formula
1	New ad, 0 results, spend < 1× target CPR	$50/day (flat starting point)
2	1 result, Tier D (below average)	$50/day
3	1 result, Tier C (at average)	$100/day
4	1 result, Tier B (above average)	min(target_daily_budget × 0.02, $500) × risk_multiplier
5	2+ results, Tier C	min(target_daily_budget × 0.015, $250) × risk_multiplier
6	2+ results, Tier B	min(target_daily_budget × 0.03, $750) × risk_multiplier
7	2+ results, Tier A	min(target_daily_budget × 0.06, $1,500) × risk_multiplier
8	Top 1-2 ads in account	min(target_daily_budget × 0.10, $3,000) × risk_multiplier

Round all band amounts to nearest $50 for clean operations.

Kill rules:

• 0 results AND spend ≥ 1× target CPR → Kill immediately
• 1 result AND spend ≥ 2× target CPR → Kill
• 2+ results AND CPR in Tier E → Drop one band, don't kill yet
• Any ad sitting at $50/day for 14+ days with no improvement → Kill (creative fatigue)

Promote-to-scale rules:

• 2+ results AND CPR in Tier A for 3+ consecutive days → Promote to Band 8
• Any ad at Band 8 for 7+ days with stable performance → Leave alone (no further increases — this is the cap)

---

Phase 4: Generate Campaign Structure Doc

Write the complete structure to outputs/campaign-structure.md:

# Campaign Structure — [Account Name or "Your Account"]
**Generated:** [Date]
**Strategy:** Manual ABO Testing + Band Management (standard)
**Funnel Type:** Lead Gen
**Target Monthly Spend:** $[from Q1]
**Target Daily Budget:** $[calculated]
**Account Average CPR:** $[from Q4]
**Target CPR:** $[from Q3]
**Risk Profile:** [1-5 from Q7]

## 1. Campaign Architecture

You will run ONE primary campaign:

**"[Account Code] — Testing — [MonthYear]"**

Type: Ad Set Budget Optimization (ABO)
Optimization event: [From Q8]
Attribution window: [From context/ad-account.md — typically 7DC 1DV for lead gen]

### Why ABO, Not CBO
You're running lead gen. Feedback loops are slow, signals are sparse, and Meta's algorithm isn't getting enough conversion data to allocate correctly. CBO will over-spend on the ad that gets cheap leads — not the ad that gets GOOD leads. ABO puts you in control of distribution so you can manually promote ads based on real results (show rate, close rate, CAC — not just cost per lead).

## 2. Ad Set Structure

**One ad set per concept.** A concept = persona + angle + format combination from your creative strategy brief.

**3 ads per ad set.** These are the 3 hook variations produced by the Script Generator (one concept, three different hooks — same body, same CTA, different opening).

**Starting budget: $50/day per ad set.**

Example ad set names (applying the naming convention from your SOP):
- `P1.2_UGC-Selfie_Ozempic-Refugee_6mo-Setup`
- `P2.1_Talking-Head_Menopause-Struggler_Free-Audit`

## 3. Your Custom Budget Bands

| Band | Condition | Daily Budget |
|------|-----------|--------------|
| 1 | New ad set, 0 results, under spend threshold | $50 |
| 2 | 1 result, CPR below average (Tier D) | $50 |
| 3 | 1 result, CPR at average (Tier C) | $100 |
| 4 | 1 result, CPR above average (Tier B) | $[calculated] |
| 5 | 2+ results, CPR at average (Tier C) | $[calculated] |
| 6 | 2+ results, CPR above average (Tier B) | $[calculated] |
| 7 | 2+ results, CPR top tier (Tier A) | $[calculated] |
| 8 | Top 1-2 ads in the account | $[calculated, capped at daily_budget × 0.10] |

## 4. Kill Rules

**Kill immediately if:**
- 0 results AND spend ≥ $[target CPR] (1× target CPR)
- 1 result AND spend ≥ $[2× target CPR]
- Any ad set at $50/day for 14+ days with no improvement

**Drop one band (don't kill yet) if:**
- 2+ results AND CPR in Tier E (>170% of account average)
- Performance degrades 2 bands in a single day

## 5. Scale Rules

**Promote up one band if:**
- Performance is at the next tier for 3+ consecutive days (3-day rolling lookback)
- Weekly review (7-day lookback) shows the ad has earned the upgrade

**Promote straight to Band 8 if:**
- 2+ results AND Tier A (top performer) for 3+ consecutive days

## 6. Management Cadence

### Daily (3 minutes, every morning)
1. Open Ads Manager → your Testing campaign
2. Filter: 3-day rolling lookback
3. Run through 4 filter views: No results / 1 result / Scale / Had delivery
4. Apply band rules mechanically — one band change per ad per day MAX
5. Use the "Date Last Edited" column to avoid double-moves

### Weekly (30 minutes, every Monday)
1. All-time lookback: kill anything over $1,000 total spend with 0 results
2. 7-day lookback: apply band rules more aggressively (promote proven winners, cut fatigued ones)
3. Pull the Daily A/B Band Management skill for a deeper diagnosis if anything looks off

## 7. Custom Filters To Set Up In Ads Manager

Create these 4 saved filter views today:

1. **Had Delivery** — Impressions > 0 (3-day lookback)
2. **No Results** — Results = 0
3. **1 Result** — Results = 1
4. **Scale** — Results ≥ 2

## 8. Custom Columns To Set Up

Name this column set "Performance Analytics":

- Ad Set Name
- Budget (Daily)
- Amount Spent
- Results (your optimization event)
- Cost Per Result
- Date Last Edited

Set this as your default column view in Ads Manager.

## 9. Conditional Formatting (Color Coding)

In Ads Manager, apply conditional formatting to the "Cost Per Result" column:
- **Green:** CPR ≤ 60% of account average (Tier A)
- **Yellow:** CPR 60-100% of account average (Tier B)
- **Orange:** CPR 100-140% of average (Tier C)
- **Red:** CPR > 140% of average or 0 results with high spend

This makes the daily 3-minute review viable.

## 10. Monthly Review & Recalibration

At the end of every month:
1. Recalculate account average CPR
2. If it's shifted by more than 20% from the value used to build these bands, re-run this skill
3. Re-examine risk tolerance — if your creative pipeline has scaled, raise risk profile; if it's shrunk, lower it

Show the user the output. Ask:

"Here's your campaign structure. Read it back. Does this match your reality — your spend, your bandwidth, your team capacity? If the bands feel too aggressive or too conservative, say so and I'll recalculate."

Iterate on band numbers if needed. Save when approved.

---

Phase 5: Tracker Sheet Template (3 min)

Generate a simple tracking sheet template and save to outputs/ad-set-tracker.md:

# Ad Set Tracker — [Account Name]

Copy this into a Google Sheet or Airtable. Update weekly (or daily if you prefer).

| Ad Set Name | Concept ID | Launched | Current Budget | Current Band | Total Spend | Total Results | CPR | Status | Date Last Edited | Notes |
|-------------|-----------|----------|----------------|--------------|-------------|---------------|-----|--------|-----------------|-------|
| P1.2_UGC-Selfie_Ozempic-Refugee_6mo-Setup | B0001 | [date] | $50 | Band 1 | | | | Testing | | |
| ... | | | | | | | | | | |

Status options: Testing / Scaling / Paused-Fatigue / Paused-Kill / Won

Tell the user:

"Drop this into a Google Sheet. Keep it updated weekly. This + the Ads Manager view + the Daily A/B Band Management skill = your complete ops stack."

---

Phase 6: Wrap & Hand-off

Save to memory/open-loops.md:

## Ads Setup Engine — Complete
Date: [date]
Artifacts:
- outputs/campaign-structure.md (custom bands, kill/scale rules, cadence)
- outputs/ad-set-tracker.md (tracker template)

Next actions:
- Build the Testing campaign in Meta using this structure
- Apply naming convention to every ad set + ad
- Set up the 4 custom filters in Ads Manager
- Set up the "Performance Analytics" custom column set
- Apply conditional formatting
- Launch first 3-5 ad sets at $50/day
- Start using the Daily A/B Band Management skill from Day 1

Tell the user:

"Your campaign structure is built. Next: go into Meta Ads Manager, build the campaign using this spec, launch your first 3-5 concepts. From tomorrow onward, your daily 3-minute ritual is the Daily A/B Band Management skill — paste your export, get your band moves, apply them.

Weekly on Monday, run the full Bottleneck Analysis skill for deeper patterns.

If you want me to walk you through the Meta Ads Manager setup step-by-step, say 'walk me through the build'."

If they want the walkthrough, provide it. Otherwise, wrap.

---

Constraints

1. Lead gen only. Direct purchase accounts get refused at Phase 1.
2. Never invent band numbers. All calculations must trace back to user inputs.
3. Starting budget is always $50/day per ad set. Non-negotiable — this is the standard modification.
4. 3 ads per ad set. Always. Hook variations from the Script Generator.
5. Push back on vague answers. If they say "I don't know my CPR", stop and send them to the KPI Tracker.
6. Round to clean numbers. $50, $100, $250, $500 — never $437 or $1,186.