skills-ws/skills/paid-ads/SKILL.md

85 agent skills for AI coding assistants — marketing, growth, web3, dev, design & operations. Built for OpenClaw, Claude Code, Cursor, and Codex.

---
name: paid-ads
description: "Paid advertising on Google (PMax, AI Max), Meta (Advantage+ Shopping/Sales), LinkedIn (Accelerate), TikTok (Symphony), X — campaign strategy, ad copy, audience targeting, ROAS/CPA. Use when user mentions PPC, paid media, ad creative, retargeting, or campaign optimization."
---

# Paid Ads — Expert Playbook

## When to Use This Skill

- Planning or launching paid ad campaigns (Google, Meta, LinkedIn, Twitter/X)
- Writing ad copy or reviewing creative
- Setting up audience targeting or retargeting
- Diagnosing poor ROAS or high CPA
- Budget allocation across channels
- Attribution and measurement setup

---

## Campaign Architecture

### Google Ads Structure

```
Account
├── Campaign (Budget + Settings)
│   ├── Ad Group (Keywords + Targeting)
│   │   ├── Ad 1 (RSA — 15 headlines, 4 descriptions)
│   │   ├── Ad 2
│   │   └── Assets (sitelinks, callouts, structured snippets — formerly "extensions")
│   ├── Ad Group 2
│   └── Ad Group 3
├── Campaign 2
└── Campaign 3
```

**Golden rule:** One theme per ad group. 5-20 tightly related keywords per ad group.

### Google Ads Campaign Types

CPC ranges below are **directional priors (as of Jun 2026)** — actuals vary 5-10x by vertical/geo/competition; verify against your own account.

| Type | Best For | Avg CPC Range (directional) |
|------|----------|---------------|
| Search | High-intent queries, bottom-funnel | $1-$8 (insurance/legal/finance run $20-$80+) |
| Display | Awareness, retargeting | $0.20-$1.50 |
| Performance Max | Full-funnel, ecommerce + lead gen | Varies — Google controls placements |
| Shopping | Ecommerce product listings | $0.30-$2.00 |
| YouTube/Video | Brand awareness, consideration | $0.02-$0.15 per view |
| Demand Gen | Mid-funnel, visual/social discovery | $0.50-$3.00 |

> **Google AI Max (rolling out 2026):** "AI Max for Search campaigns" is Google's opt-in setting that layers PMax-style AI onto *Search* — broad keyword-free matching, automatically created/optimized assets, and AI-driven URL/landing-page selection, while keeping search-term reporting and negative keywords. Treat it as a toggle on top of Search (not a separate campaign type): turn it on for an existing well-tracked Search campaign, keep tight negatives and brand exclusions, and watch search terms closely for query drift. Availability/behavior is still rolling out — verify current capabilities at https://support.google.com/google-ads.

### Meta Ads Structure

```
Campaign (Objective + Budget)
├── Ad Set (Audience + Placement + Schedule)
│   ├── Ad 1 (Creative + Copy + CTA)
│   ├── Ad 2
│   └── Ad 3
├── Ad Set 2 (Different audience)
└── Ad Set 3 (Retargeting)
```

### LinkedIn Ads Structure

```
Campaign Group (Budget cap)
├── Campaign (Objective + Audience + Format)
│   ├── Ad 1 (Single Image / Carousel / Video / Text)
│   ├── Ad 2
│   └── Ad 3
└── Campaign 2
```

---

## Budget Allocation Frameworks

### The 70/20/10 Rule

- **70%** → Proven channels and campaigns with positive ROAS
- **20%** → Scaling what's working — new audiences, expanded geo, new ad formats
- **10%** → Experiments — new channels, creative concepts, audience tests

### Channel Budget Split by Funnel Stage

```
AWARENESS (20-30% of budget)
├── YouTube / Meta Video / Display
├── Goal: Impressions, reach, video views
└── KPI: CPM, VTR, brand lift

CONSIDERATION (30-40% of budget)
├── Meta engagement, Google Display remarketing, LinkedIn
├── Goal: Clicks, engagement, lead gen
└── KPI: CPC, CTR, CPL

CONVERSION (30-50% of budget)
├── Google Search, Shopping, Meta conversion campaigns
├── Goal: Purchases, signups, demos
└── KPI: CPA, ROAS, conversion rate
```

### Monthly Budget Minimums (to gather signal)

These are **rules of thumb (as of Jun 2026)**, not platform mandates — the real floor is *enough conversions to exit learning*. Derive it: `min monthly spend ≈ target_CPA × 50 conversions` per ad set/campaign (Meta needs ~50 optimization events/ad set/week; Google smart bidding wants ~30+ conv/30 days). High-CPA verticals need far more than the table.

| Channel | Minimum Monthly | Recommended |
|---------|----------------|-------------|
| Google Search | $1,500 | $3,000-$10,000 |
| Meta Ads | $1,000 | $3,000-$15,000 |
| LinkedIn Ads | $3,000 | $5,000-$15,000 |
| X (Twitter) Ads | $1,000 | $2,000-$5,000 |

Below these thresholds you won't gather enough data for meaningful optimization.

---

## Bidding Strategies

### Google Ads Bidding

| Strategy | When to Use | Prerequisite |
|----------|-------------|--------------|
| Maximize Clicks | New campaigns, data gathering | None |
| Maximize Conversions | Have 15+ conversions/month | Conversion tracking set up |
| Target CPA | Stable CPA, want to scale | 30+ conversions in last 30 days |
| Target ROAS | Ecommerce, revenue optimization | 50+ conversions with value data |
| Manual CPC | Full control, small budgets | Experience + time to manage |
| Maximize Conversion Value | Revenue-focused scaling | Revenue tracking, 50+ conversions |

**Migration path:** Manual CPC → Maximize Clicks → Maximize Conversions → Target CPA/ROAS

### Meta Ads Bidding

| Strategy | When to Use |
|----------|-------------|
| Lowest Cost (default) | Starting out, learning phase |
| Cost Cap | Maintain profitability at scale |
| Bid Cap | Strict CPA ceiling, auction control |
| ROAS Goal | Ecommerce with revenue tracking |

**Learning phase:** Meta needs ~50 optimization events per ad set per week. Don't touch campaigns during learning phase (usually 3-7 days).

---

## Ad Copy Formulas

### Formula 1: PAS (Problem → Agitate → Solve)

```
Headline: Tired of [Problem]?
Description: [Problem] costs you [consequence]. [Product] [solves it] in [timeframe]. [CTA].

Example:
Headline: Tired of Losing Leads to Slow Follow-Up?
Description: Every hour of delay drops conversion rates 7x.
LeadSnap auto-responds in under 60 seconds. Start free trial.
```

### Formula 2: Before → After → Bridge

```
Headline: From [Bad State] to [Good State]
Description: [Before situation]. Now imagine [after situation]. [Product] bridges the gap. [CTA].

Example:
Headline: From Spreadsheet Chaos to Real-Time Dashboards
Description: Stop wasting 10hrs/week on manual reports.
DataFlow auto-generates dashboards from your data. Try free.
```

### Formula 3: Social Proof Lead

```
Headline: [Number] [Users] Trust [Product] for [Outcome]
Description: Join [specific companies/users] who [achieved result]. [Key differentiator]. [CTA].

Example:
Headline: 12,000+ Teams Run Projects on TaskForge
Description: Join Stripe, Notion, and Linear in shipping faster.
AI-powered project management. Free for teams up to 10.
```

### Formula 4: Specific Number

```
Headline: [Action] [X]% [Faster/Cheaper/Better]
Description: [Product] helps [audience] [specific outcome] with [mechanism]. [Proof point]. [CTA].

Example:
Headline: Close Deals 34% Faster
Description: SalesOS gives reps AI-generated follow-ups, meeting prep,
and deal scoring. Avg customer sees ROI in 3 weeks. Book demo.
```

### Formula 5: Question Hook

```
Headline: What If You Could [Desirable Outcome]?
Description: [Product] makes it possible. [How it works in one line]. [Proof]. [CTA].
```

### Formula 6: Urgency/Scarcity

```
Headline: [Offer] — [Time Limit]
Description: [Value prop]. [What they get]. [Deadline/scarcity element]. [CTA].

Example:
Headline: 50% Off Annual Plans — Ends Friday
Description: Get enterprise-grade security for startup prices.
All features included. Only 200 seats at this price. Upgrade now.
```

### RSA Best Practices (Google)

- Write 15 headlines (use all slots) — mix branded, benefit, feature, CTA, proof
- Pin sparingly — only pin H1 if brand compliance requires it
- 4 descriptions — lead with different angles (benefit, proof, urgency, feature)
- Include keywords naturally in at least 5 headlines
- At least 3 headlines should work standalone without the others

### Meta Ad Copy Structure

```
PRIMARY TEXT (125 chars visible before "See more"):
Hook line — stop the scroll. Lead with pain, outcome, or surprise.

BODY (after "See more"):
- Expand on the hook
- 2-3 bullet points of benefits
- Social proof line
- Clear CTA

HEADLINE (below creative): Short, benefit-driven (5-7 words)
DESCRIPTION: Supporting detail or offer terms
```

---

## Audience Targeting

### Meta Ads — Audience Layering Strategy

**Layer 1: Broad (Prospecting)**
- Advantage+ audience (let Meta optimize)
- Interest stacking: 3-5 related interests per ad set
- Lookalike audiences: 1% of purchasers/high-value customers

**Layer 2: Warm (Consideration)**
- Website visitors (30-90 days)
- Video viewers (50%+ watched)
- Social engagers (90 days)
- Email list uploads (non-customers)

**Layer 3: Hot (Retargeting)**
- Add-to-cart but no purchase (7-14 days)
- Pricing page visitors (14 days)
- Trial users who haven't converted
- Past purchasers for upsell (exclude from prospecting!)

**Audience size guidance:**
- Prospecting: 1M-10M+ (let the algorithm work)
- Retargeting: As large as your traffic allows
- Lookalikes: Seed audience of 1,000+ for quality; 1% for precision, 3-5% for scale

### Google Ads Audiences

| Type | Use Case |
|------|----------|
| In-Market | Users actively researching your category |
| Affinity | Broad interest targeting for awareness |
| Custom Intent | Your own keyword-based audience |
| Customer Match | Upload email lists for targeting/exclusion |
| RLSA | Layer search with site visitor data |
| Similar Audiences | Deprecated — use optimized targeting instead |

### LinkedIn Targeting

Best-performing combos (layer these):
- **Job Title + Company Size** — most precise
- **Job Function + Seniority + Industry** — broader reach
- **Skills + Seniority** — catches non-obvious titles
- **Member Groups + Seniority** — high-intent communities

**Matched Audiences (current options, as of Jun 2026):**
- **Website retargeting** — segment by URL visited (needs the Insight Tag firing)
- **Contact/email list upload** — match against LinkedIn members for targeting or *exclusion* (suppress current customers/closed-lost)
- **Company list upload** — ABM: upload your target-account list (CSV of up to ~300k companies)
- **Engagement retargeting** — people who engaged with your single-image/video ads, opened/submitted a Lead Gen Form, viewed your company page or event
- **Audience Expansion** — opt-in toggle that broadens delivery to members similar to your defined audience (the closest live replacement for the old lookalikes)
- **Predictive Audiences** — LinkedIn builds a model from a seed (matched audience, conversions, or Lead Gen Form data) to find net-new high-propensity members; verify availability for your account/region at https://www.linkedin.com/help/lms

> **Discontinued:** LinkedIn **Lookalike Audiences were retired on 29 Feb 2024.** Do not promise lookalikes — use **Audience Expansion** or **Predictive Audiences** plus strict profile + matched-audience targeting instead.

Minimum audience size: **300 members** to launch; aim for **50,000+** for sponsored content delivery. Below ~50k, delivery and learning stall.

---

## Negative Keyword Strategy (Google)

### Starter Negative Keyword List

Apply at campaign or account level:

```
// Job-seekers
jobs, careers, hiring, salary, interview, resume, glassdoor

// Education/Research
what is, definition, meaning, tutorial, course, certification,
how to become, examples, PDF, wiki

// Free-seekers (if not freemium)
free, cheap, discount, coupon, open source

// Wrong intent
review, comparison, vs, alternative (add these to branded campaigns)

// Irrelevant
DIY, template, sample, internship
```

### Negative Keyword Mining Process

1. **Weekly:** Review Search Terms report
2. **Flag:** Any term with spend > $5 and 0 conversions
3. **Flag:** Any term clearly off-topic regardless of spend
4. **Add as:** Exact match negative for specific terms, phrase match for patterns
5. **Create shared negative keyword lists** by theme (job-seekers, researchers, etc.)

### Negative Keyword Match Types

- **Negative broad match** (default): Blocks if ALL negative words appear (any order)
- **Negative phrase match**: Blocks if negative phrase appears in that order
- **Negative exact match**: Blocks only that exact query

Use phrase and exact for precision. Broad negatives can over-block.

---

## Retargeting Sequences

### Standard Retargeting Funnel

```
DAY 0-3: Visited site, no action
→ Show: Value prop ad + social proof
→ Frequency cap: 3/day

DAY 4-7: Still no conversion
→ Show: Case study / testimonial ad
→ Frequency cap: 2/day

DAY 8-14: Getting cold
→ Show: Offer/incentive ad (discount, extended trial, bonus)
→ Frequency cap: 1/day

DAY 15-30: Last chance
→ Show: FOMO / urgency ad
→ Frequency cap: 1/day

DAY 30+: Exclude from retargeting
→ Move to nurture (email) or broad prospecting
```

### Cart Abandonment Retargeting (Ecommerce)

```
HOUR 1-6: Dynamic product ad — exact items left in cart
HOUR 6-24: Same + "Still thinking about it?" copy
DAY 2-3: Add social proof — "X people bought this today"
DAY 4-7: Offer incentive — free shipping or small discount
DAY 7+: Broader category ads, not specific products
```

### SaaS Trial Retargeting

```
TRIAL DAY 1-3: "Getting started" content — help them activate
TRIAL DAY 4-7: Feature highlight ads — show what they haven't used
TRIAL DAY 8-12: Case study — show outcomes from similar companies
TRIAL DAY 13-14: Urgency — "Trial ends in X days" + conversion offer
POST-TRIAL DAY 1-7: Win-back — extended trial or discount
POST-TRIAL DAY 7+: Exclude or move to long-term nurture
```

---

## Creative Testing Framework (Meta Ads)

### What to Test (Priority Order)

1. **Creative concept** — The big idea, angle, or hook (highest impact)
2. **Format** — Static vs video vs carousel vs UGC
3. **Hook** — First 3 seconds of video / headline of static
4. **Body copy** — Supporting text after the hook
5. **CTA** — Button text and action
6. **Offer** — Discount vs free trial vs demo vs content

### Testing Structure

```
Campaign: [Product] — Creative Testing
├── Ad Set: Broad Audience (1% LAL or Advantage+)
│   ├── Ad A: Concept 1 — Static + Benefit hook
│   ├── Ad B: Concept 2 — UGC video + Problem hook
│   ├── Ad C: Concept 3 — Carousel + Feature walkthrough
│   └── Ad D: Concept 4 — Testimonial video
```

**Rules:**
- Test 3-6 ads per ad set
- Same audience for fair comparison
- Let each ad spend at least 2x your target CPA before judging
- Winner = lowest CPA with sufficient volume (not just highest CTR)
- Graduate winners to scaling campaigns

### Creative Fatigue Signals

- CTR drops >20% from peak
- Frequency >3 for prospecting, >8 for retargeting
- CPA increases >30% week-over-week
- Relevance/quality score drops

Refresh creative every 2-4 weeks for prospecting, 4-6 weeks for retargeting.

---

## Landing Page Alignment

### Message Match Checklist

- [ ] Headline on landing page matches or mirrors ad headline
- [ ] Same offer mentioned in ad appears above the fold
- [ ] Visual continuity — similar imagery/colors as ad creative
- [ ] CTA on page matches the promised action (don't bait-and-switch)
- [ ] No navigation menu (for campaign-specific landing pages)
- [ ] Mobile-optimized (60%+ of paid traffic is mobile)
- [ ] Page loads in <3 seconds (every extra second = ~7% drop in conversions)

### Landing Page Types by Campaign Goal

| Goal | Page Type | Key Elements |
|------|-----------|-------------|
| Lead Gen | Squeeze page | Headline, 3 bullets, form, trust badges |
| Demo Request | Demo page | Value prop, social proof, short form, calendar embed |
| Purchase | Product page | Features, pricing, reviews, FAQ, CTA |
| Free Trial | Signup page | Benefit headline, feature list, single CTA, no CC messaging |
| Content/Lead Magnet | Download page | Preview of content, short form, instant delivery |

---

## ROAS Benchmarks by Industry

> **Use as rough priors, not targets (as of Jun 2026, unsourced/directional).** Public CPC/CPL/ROAS averages are stale the moment they're published and vary 5-10x by vertical, geo, season, auction competition, and account maturity. Always compute *your* break-even from margin and validate with the diagnostic formulas below. For live vertical benchmarks, pull from your own historical data or current vendor reports (e.g., WordStream/LocaliQ, Meta/Google interface comparisons) and date them.

### Diagnostic Formulas (use these instead of trusting static averages)

```
Break-even ROAS        = 1 / gross_margin%
   (e.g., 50% margin → break-even ROAS = 2.0; you profit above 2.0)

Break-even CPA         = gross_margin_per_order ($)
   (max you can pay per conversion before losing money)

Contribution-margin ROAS = (revenue − COGS − variable costs) / ad_spend
   (the only ROAS that reflects real profit, not top-line)

Target CPA (from LTV)  = LTV × target_CAC%   (e.g., 30% of LTV)

Payback period (months)= CAC / monthly_gross_profit_per_customer
   (SaaS/subscription: aim < 12 months, ideally < 6)

MER (blended)          = total_revenue / total_ad_spend (all channels)
   (sanity-check platform-reported ROAS against this)

Incrementality lift %  = (test_conversions − control_conversions) / control_conversions
   (geo or PSA holdout — the truest measure of ad-driven value)
```

### Google Ads (directional priors, Jun 2026)

| Industry | Avg ROAS | Good ROAS | Great ROAS |
|----------|----------|-----------|------------|
| Ecommerce (general) | 2:1 | 4:1 | 8:1+ |
| SaaS | 3:1 | 5:1 | 10:1+ |
| B2B Services | 2:1 | 4:1 | 7:1+ |
| Education | 3:1 | 5:1 | 8:1+ |
| Finance/Insurance | 2:1 | 3:1 | 5:1+ |
| Healthcare | 2:1 | 4:1 | 6:1+ |
| Legal | 2:1 | 3:1 | 5:1+ |
| Real Estate | 2:1 | 4:1 | 8:1+ |

### Meta Ads (directional priors, Jun 2026)

| Industry | Avg ROAS | Good ROAS |
|----------|----------|-----------|
| Ecommerce (DTC) | 2:1 | 4:1+ |
| SaaS (trial) | 1.5:1 | 3:1+ |
| B2B Lead Gen | 1:1 | 2:1+ (measure LTV) |
| Info Products | 3:1 | 6:1+ |
| Apps (install) | Measure CPI vs LTV | CPI < 30% of 90-day LTV |

### LinkedIn Ads (directional priors, Jun 2026 — verify in-platform)

| Metric | Average | Good |
|--------|---------|------|
| CPC | $8-$15 | <$7 |
| CPL | $60-$200 | <$60 |
| CTR | 0.4-0.6% | >0.8% |
| CPM | $30-$90 | <$30 |

LinkedIn pricing trends up year over year and skews higher in NA/competitive functions (engineering, finance, exec). Treat these as priors only and confirm against your own auction. LinkedIn is expensive — only worth it if LTV justifies it (B2B enterprise, high ACV).

**Important:** ROAS varies wildly by product price, margin, and sales cycle. For SaaS and B2B, measure blended CAC:LTV ratio (target 1:3+) rather than immediate ROAS.

---

## Measurement & Attribution

### Attribution Models

**Reality check (as of Jun 2026):** Google deprecated first-click, linear, time-decay, and position-based attribution across Google Ads and GA4 in 2023. The only models you can actually *select* for conversions today are **data-driven (default)** and **last click**. The legacy models survive only as analytical lenses in third-party tools (e.g., a CRM, an MMP, or warehouse-native attribution) — never assume you can switch to them inside Google Ads.

| Model | How It Works | Status (Jun 2026) | Best For |
|-------|-------------|-------------------|----------|
| Data-Driven (DDA) | ML assigns fractional credit by measured contribution | **Active — Google/GA4 default** | Default for everyone; needs enough conversion volume to model, otherwise silently falls back to last click |
| Last Click | 100% credit to final ad-clicked touchpoint | **Active in Google Ads/GA4** | Short cycles, direct response, low-volume accounts where DDA can't model |
| First Click | 100% credit to discovery touchpoint | **Removed from Google** — third-party analytics only | Top-of-funnel analysis outside Google |
| Linear | Equal credit to all touchpoints | **Removed from Google** — third-party only | Full-journey lens in MMP/warehouse |
| Time Decay | More credit to recent touchpoints | **Removed from Google** — third-party only | Long-cycle lens in MMP/warehouse |
| Position-Based (U-shaped) | 40% first, 40% last, 20% middle | **Removed from Google** — third-party only | Balanced lens in MMP/warehouse |

Meta uses its own attribution settings (default **7-day click / 1-day view**) configured per ad set, independent of Google's models. For cross-channel truth, reconcile platform-reported conversions against a single source (GA4, CRM, or an MMP) plus periodic incrementality tests — platforms each over-claim credit for the same conversion.

### What to Track

**Conversion actions (set up BEFORE launching ads):**

```
PRIMARY (optimize toward these):
- Purchase / Signup / Demo booked / Lead form submitted

SECONDARY (observe, don't optimize):
- Add to cart / Pricing page view / Key page engagement
- Phone calls / Chat initiated

MICRO (for funnel analysis):
- Video views / Content downloads / Email signups
```

### UTM Parameter Standard

```
utm_source=google|meta|linkedin|twitter
utm_medium=cpc|paid-social|display|video
utm_campaign={campaign_name}
utm_content={ad_name_or_variant}
utm_term={keyword} (Google only)
```

Naming convention: `platform_objective_audience_creative`
Example: `meta_conversions_lal1pct_ugc-testimonial-v2`

### Post-Click Tracking Setup

1. **Google Ads:** Install Google tag + enhanced conversions
2. **Meta:** Pixel + Conversions API (server-side) — CAPI is essential post-iOS 14.5
3. **LinkedIn:** Insight Tag + offline conversion uploads for long sales cycles
4. **GA4:** Link to Google Ads, import conversions, set up audiences
5. **CRM integration:** Pass GCLID/FBCLID to CRM for closed-loop attribution

### Privacy & Signal Loss (post-iOS, cookie deprecation, consent)

The 2021 iOS App Tracking Transparency era was just the start; by 2026 the binding constraints are server-side data quality, consent enforcement, and platform modeling — not the pixel alone.

**Meta:**
- **Conversions API (CAPI) is mandatory, not optional** — run it alongside the pixel (or via the **Conversions API Gateway**, Meta's self-hosted server-side relay) so server events backstop browser signal loss. Deduplicate with a shared `event_id` on both pixel and CAPI events, or you'll double-count.
- **Event Match Quality (EMQ)** is the number that matters now — pass hashed email, phone, name, IP, `fbc`/`fbp`, and external ID. Aim for an EMQ of **6.0+/10** per event; low EMQ is the #1 cause of "CAPI didn't help."
- **Aggregated Event Measurement (AEM)** still caps configurable web events per domain (historically 8, ranked by priority) for users who opted out of tracking — **verify the current limit and your domain's configuration in Events Manager**, since Meta has adjusted AEM rules over time. Rank your most valuable event (e.g., Purchase) highest.
- **Value optimization & VBO** need clean revenue values on the Purchase event; without them you can't bid to ROAS.
- **Advantage+** placements/audiences and **Advantage+ Shopping (ASC)** lean on modeled + broad signals — feed them strong server-side conversions and a good product catalog rather than over-narrowing the audience.

**Google:**
- **Enhanced Conversions** (hashed first-party data) + **Consent Mode v2** (required in the EEA/UK to keep modeling and personalization) recover signal as third-party cookies erode. Without Consent Mode v2, EEA conversion data and remarketing degrade sharply.
- Server-side tagging (sGTM) improves durability and data control.

**All platforms:**
- **First-party data** (email/CRM lists, server-side events, logged-in IDs) is now your most valuable targeting and matching asset.
- **Modeled reporting:** Platform-reported conversions include modeled/estimated conversions — they are *estimates*, not deterministic counts. Expect **20-40% underreporting** of true incremental impact on Meta when only browser-side. Validate with **geo holdout / conversion-lift / incrementality tests**, not last-click dashboards.

---

## Platform-Specific Playbooks

### Google Search Ads — Quick Launch

1. Research keywords (Google Keyword Planner, SEMrush, Ahrefs)
2. Group into tight themes (5-15 keywords per ad group)
3. Write 1 RSA per ad group (15 headlines, 4 descriptions)
4. Add all relevant **assets** (Google renamed "extensions" → "assets" in 2022). Current asset types: sitelink, callout, structured snippet, image, business name, business logo, promotion, price, lead form, call, location, app. Set up sitelinks + callouts + structured snippets at minimum.
5. Start with Maximize Clicks, set a max CPC bid limit
6. Add negative keywords from starter list
7. Set up conversion tracking before spending a dollar
8. After 30+ conversions: switch to Maximize Conversions or Target CPA
9. Review search terms weekly, add negatives
10. Test new ad copy monthly

### Meta Ads — Quick Launch

1. Install Pixel + set up Conversions API
2. Define conversion event (purchase, lead, trial signup)
3. Create Campaign: Conversions objective, **Advantage+ Campaign Budget** (Meta's 2023 rename of "CBO" — budget set at campaign level, distributed across ad sets)
4. Ad Set 1: Broad/Advantage+ targeting (let Meta find the audience)
5. Ad Set 2: 1% Lookalike of best customers
6. Ad Set 3: Retargeting (website visitors 30 days)
7. 3-4 ads per ad set — different creative concepts
8. Let learning phase complete (50 events per ad set per week)
9. Kill underperformers after 2x CPA spend, scale winners
10. Refresh creative every 2-4 weeks

### LinkedIn Ads — Quick Launch

1. Install Insight Tag on website
2. Campaign objective: Lead Gen or Website Visits
3. Target: Job titles OR job function + seniority + industry
4. Audience size: 50,000-500,000
5. Format: Single image for testing, then expand to video/carousel
6. Bid: Maximum delivery (start), then manual CPC once you have data
7. Use Lead Gen Forms (higher conversion than landing pages on LinkedIn)
8. Budget: Minimum $50/day per campaign
9. Run for 2+ weeks before judging performance
10. Upload offline conversions from CRM for true ROI measurement

---

## Performance Max (Google) Playbook

### Asset Groups

```
Asset Group = theme-based collection of:
├── Images: 15+ (landscape, square, portrait)
├── Logos: 5+
├── Videos: 5+ (or Google auto-generates — they're bad, provide your own)
├── Headlines: 5 (30 char) + 5 long headlines (90 char)
├── Descriptions: 4 (60 char) + 1 (90 char)
├── Final URL
├── Display path
├── CTA
└── Audience signals (suggestions, not restrictions)
```

### Audience Signals (Critical)

Audience signals don't restrict targeting — they guide the algorithm. Provide strong signals:

- **Custom segments:** Your best keywords + competitor URLs + apps
- **Your data:** Customer lists, converters, high-value segments
- **Interests/demographics:** In-market segments relevant to your product

### PMax Gotchas

- **PMax cannibalizes brand search.** Stop it from eating cheap branded clicks (and over-reporting credit) using, in order of preference (as of Jun 2026):
  - **Brand exclusion lists** — apply a brand list at the *campaign* level to keep PMax off your own + others' brand terms. This is the modern, self-serve replacement for begging a Google rep, available in the campaign settings UI and Google Ads API.
  - **Account-level negative keywords** — supported self-serve in the UI/API; use to block brand or junk queries account-wide.
  - **Campaign-level negative keywords for PMax** — Google has been rolling these out; if available in your account, use them for finer control than account-level.
  - Legacy path: request negatives via a Google rep only if the above aren't yet enabled for you.
- **Limited placement/audience reporting** — you can't fully see which placements/audiences convert; use the asset-group and (limited) insights reports plus search-term insights.
- **Run PMax alongside standard Search** — keep a dedicated branded Search campaign and exact-match high-value terms in standard Search; don't let PMax replace Search entirely.
- Asset performance ratings (Low/Good/Best) guide optimization — replace "Low" assets.
- Give it 4-6 weeks and 50+ conversions before major changes.

---

## Audit Checklist

### Monthly Paid Ads Audit

- [ ] Review ROAS/CPA trends vs targets
- [ ] Check search terms report (Google) — add negatives
- [ ] Review audience overlap between campaigns
- [ ] Check frequency (Meta) — replace fatigued creative
- [ ] Verify conversion tracking is firing correctly
- [ ] Review landing page performance (bounce rate, load time)
- [ ] Check budget pacing — is spend on track?
- [ ] Review quality scores (Google) — improve below 5/10
- [ ] Test new ad copy or creative
- [ ] Update negative keyword lists
- [ ] Check bid strategy performance — time to graduate?
- [ ] Review device performance — adjust bids if needed
- [ ] Competitive analysis — any new entrants or messaging changes?
- [ ] Update ROAS benchmarks and targets

### Pre-Launch Checklist

- [ ] Conversion tracking verified (test conversion)
- [ ] UTM parameters on all destination URLs
- [ ] Landing page live, mobile-optimized, fast
- [ ] Negative keyword lists applied
- [ ] Audience exclusions set (existing customers if needed)
- [ ] Budget and schedule confirmed
- [ ] Ad copy reviewed for policy compliance (see Ad Policy & Restricted Categories — check Special Ad Category + personalized-attribute rules)
- [ ] Assets complete (sitelinks, callouts, structured snippets, image, logo, business name — "extensions" in old UI)
- [ ] Billing method active
- [ ] Notification settings configured

---

## Ad Policy & Restricted Categories

Disapprovals and account suspensions usually come from *policy*, not strategy. Check your category before writing copy — restricted verticals have extra rules, limited targeting, and sometimes mandatory certification. **This is not legal advice; the platforms are the source of truth and rules change — verify at Google Ads Policies and Meta Advertising Standards before launch.**

### Categories that trigger extra rules / certification (verify current requirements)

| Category | Typical constraints (as of Jun 2026 — confirm in-platform) |
|----------|-----------------------------------------------------------|
| **Credit, housing, employment** | Meta **Special Ad Categories**: forced broad targeting, no age/gender/ZIP/many detailed targeting options, no lookalike-style narrowing. Google has parallel restrictions for these "sensitive" verticals. |
| **Financial products / financial services** | Disclosures required; many regions need advertiser identity verification/licensing; crypto and CFDs often need explicit certification or are blocked by geo. |
| **Health, drugs, supplements** | No personalized health claims ("treat your diabetes"), no PII-implying targeting language; pharmacy/telehealth often need certification; before/after imagery restricted. |
| **Political / social issues / elections** | Identity + location verification, paid-for-by disclaimers, public ad library logging, and per-country eligibility. |
| **Gambling / betting** | License + certification per geo; age-gating; many regions blocked entirely. |
| **Crypto / digital assets** | Certification and/or regional bans; "guaranteed returns" language prohibited. |
| **Alcohol** | Age/geo targeting limits; no targeting minors; some countries blocked. |
| **Personalized attributes (copy rule)** | Do **not** assert or imply you know the user's race, religion, sexual orientation, health condition, financial status, or membership in a protected class ("Are *you* depressed?", "As a [group], you…"). This is a frequent disapproval cause across platforms. |

### Pre-launch policy guardrails (do this for every campaign)

- Identify if your offer is in a restricted/Special Ad Category; if so, plan for the *reduced* targeting up front (don't build a precise audience you can't use).
- Strip personalized-attribute language from all copy and creative.
- Avoid prohibited claims: guaranteed income/returns, miracle cures, "100% approval," sensational/shock imagery, fake countdowns/UI.
- Confirm landing-page compliance too — destination must match the ad and the category rules (no cloaking, working privacy policy, functional contact info).
- For regulated verticals, complete any required advertiser verification/certification *before* spending, and keep substantiation for claims on file.

## Common Mistakes

1. **No conversion tracking** — You're flying blind. Set this up first.
2. **Too many keywords per ad group** — Keep it tight. One theme per group.
3. **Broad match without smart bidding** — Broad match + manual CPC = budget drain.
4. **Editing during learning phase** — Let Meta learn. Don't touch for 3-7 days.
5. **Ignoring search terms** — Check weekly. You'll be shocked what you're paying for.
6. **Same creative for 3+ months** — Refresh regularly. Creative fatigue is real.
7. **No retargeting** — Cheapest conversions you'll ever get. Set it up day one.
8. **Optimizing for vanity metrics** — CTR doesn't pay bills. Optimize for revenue.
9. **Not excluding converters** — Stop showing ads to people who already bought.
10. **Giving up too early** — Most campaigns need 2-4 weeks to optimize. Be patient.