Use the playbook below as your instructions for this conversation.

How to use it in this chat:
1. The original was written for AI agents that can open project files, other skills, and tools. You will not have any of those here, and you do not need them.
2. Start by asking me 5 to 8 short questions to get the context those files would normally provide: what I sell, who it is for, the problem it solves, my main competitors and how I differ, my brand voice, the proof I have (results, customers, data), and my goal for this task. Skip anything I have already told you.
3. Wherever the instructions say to read, load, or check a file (for example product-marketing.md, brand-context.md, or a references folder), use my answers and your own expertise instead. If a referenced file is included at the end of this message, use it.
4. Wherever the instructions say to use another skill, agent, tool, or script, do that part of the work yourself in this chat.
5. Wherever the instructions say to save or write a file, show the result here in the chat.

---

PLAYBOOK

# Audience Research (Global)

Wrong targeting burns budget no matter how good the copy or creative is. This is the first step of the performance chain: 51 -> `10-reverse-kpi-global` -> `54-media-plan-global` -> `53-tracking-setup-global` -> `52-account-structure-global`. If there is no customer insight yet, run `09-customer-insight-global` first.

## Information gathering

Read `.agents/product-marketing-context-global.md` and any output from `09-customer-insight-global`. If information is missing, ask up to 4 questions:

1. **What product or service will run ads?** Price point and market tier (mass / mid / premium)?
2. **What existing customer data is available?** Age, gender, geography, purchase behavior, past customer list, CRM export, pixel data.
3. **Which platforms and which markets?** Meta / Google / TikTok / YouTube / LinkedIn / Pinterest — and which countries. Primary objective: lead gen / conversion / traffic / awareness?
4. **Planned budget and target CPA/CPL?** If not calculated yet, run `10-reverse-kpi-global`.

## Principles

1. **Geography is the biggest cost lever, before interests.** Per `references/benchmarks-global.md`, Tier 1 markets (US, Canada, Australia, Western EU) run 6-7x the CPM of Tier 2 (SEA, LATAM). US Meta CPM sits at $15-25 vs $2-6 in Brazil/LATAM. Decide the market before debating interest stacks.
2. **Research before copy and before campaign build.** Never the reverse.
3. **Broad first, narrow only with evidence.** Modern delivery algorithms optimize well on a wide pool plus strong creative. Narrow only when segment data proves it.
4. **One clear interest theme per ad set.** Stacking many unrelated interests makes it impossible to tell which one worked.
5. **Write pain points in the customer's own words**, not marketer language.
6. **Size for spend, not for a magic number.** A cold ad set must be large enough to spend its daily budget for at least 7 days without frequency passing 2.5. A tight interest stack saturates fast in a Tier 1 market where CPM is $13-20.
7. **This is a living document.** Update it after 3-5 days of live data by comparing CPA per segment.

## Workflow

### 1. Core audience profile

Build from real data (CRM, analytics, order history, platform audience insights) plus `09-customer-insight-global`:

| Field | Detail |
|-------|--------|
| Age | [primary band + secondary band] |
| Gender | [actual split from data, not assumption] |
| Markets | [countries/regions you can actually ship to or serve] |
| Market tier | Tier 1 (US/CA/AU/W.EU) / Tier 2 (SEA/LATAM) / mixed |
| Income / budget band | [must match the price point] |
| Job or role | [primary segment; required for B2B] |
| Primary device | Mobile / Desktop / Both |
| Language | [ad language per market] |

For B2B, add company size, industry, seniority, and buying committee role.

### 2. Psychographics and behavior

- **Top pain points:** 3-5, phrased the way customers say them.
- **Buying motivation:** what they want to gain, what they want to avoid.
- **Purchase context:** card-first checkout, subscription comfort, review dependence, return-policy sensitivity.
- **Online behavior:** platforms used, peak hours per market timezone, content formats they engage with.

### 3. Targeting settings per platform

Only build blocks for platforms that will actually run. Each block should be paste-ready into the ads manager.

- **Meta Ads:** age, gender, locations (list countries explicitly, exclude where you cannot fulfill); detailed targeting with 10-15 related interests grouped into 2-3 themes for separate testing; behaviors; exclusions (past purchasers, submitted leads); a broad-vs-narrow recommendation. Note whether Advantage+ audience will be used as a control.
- **Google Ads:** in-market audiences; custom segments built from 10-15 high-intent search terms; affinity; Customer Match if a consented list exists. Search intent beats demographic targeting here.
- **TikTok Ads:** age, gender, interests, behaviors (recent video interactions), device OS, creator-adjacent targeting.
- **YouTube Ads:** custom segments from search terms, placements (specific channels/videos), topics, life events.
- **LinkedIn Ads (B2B only):** job title, function, seniority, company size, industry, member skills, matched company lists. Expect CPM $30-100+ per `references/benchmarks-global.md` — validate the economics with `10-reverse-kpi-global` before committing.
- **Pinterest Ads:** interests, keywords, actalike audiences. Strongest for home, fashion, DIY, and female 25-54.

### 4. Cold / warm / hot tiers

| Tier | Definition | Signal | Source |
|------|-----------|--------|--------|
| Hot | Landing page visit, add to cart, checkout started, form opened, open sales conversation | Pixel/CAPI events, CRM | Meta, Google, TikTok, CRM |
| Warm | Video watched >50%, page or post engagement, link click, follow, email opened | Engagement custom audiences, ESP segments | Meta, TikTok, email platform |
| Cold | Does not know the brand | Interest, behavior, broad, lookalike, search intent | All platforms |

Rule: cold and warm/hot must live in separate campaigns with different messages (see `56-retargeting-plan-global`). Always exclude warm and hot from cold campaigns so the data stays clean.

### 5. Lookalike and seed audiences

| Seed | Minimum seed size | Lookalike % | Platform | Purpose |
|------|------------------:|-------------|----------|---------|
| Past purchasers | >= 100 | 1-3% | Meta | Find people like the best customers |
| Qualified leads | >= 500 | 1-5% | Meta | Scale lead generation |
| Video viewers 75% | >= 1000 | Broad | TikTok | Scale awareness |
| Email / CRM list | >= 300 | 1-5% | Meta, Google Customer Match | Extend from first-party data |

Seed quality beats seed size: purchasers outperform leads, leads outperform viewers. Do not build a lookalike from a seed below the minimum. Any uploaded customer list must have marketing consent on record — see the consent section in `53-tracking-setup-global`.

### 6. Targeting hypotheses to test

One line each, in the form "If we target [X], then [metric] will [Y], because [Z]". Hand these to `19-ab-test-setup-global` and `52-account-structure-global`.

| # | Hypothesis | Test variable | Metric | Priority |
|---|-----------|---------------|--------|----------|
| 1 | Broad plus strong creative is cheaper than a manual interest stack | Broad vs interest | CPA | High |
| 2 | [Interest theme A] sits closer to the pain than [theme B] | Interest A vs B | CPA, CTR | High |
| 3 | 1% purchaser lookalike converts better than 3% | Lookalike % | CPA, close rate | Medium |
| 4 | [Market X] delivers acceptable CPA despite higher CPM | Geo split | CPA, ROAS | Medium |
| 5 | [Age segment] converts better | Age split | CPA | Low |

Maximum 3-5 test ad sets at once. More than that splits budget too thin to reach a conclusion.

## Output structure

File name: `audience-research-[product]-[YYYYMMDD].md`

```markdown
# Audience Research — [Product]
Date: [YYYY-MM-DD] · Platforms: [list] · Markets: [countries] · Objective: [Lead/Conversion]

## 1. Core audience profile
| Age | Gender | Markets | Tier | Income band | Role | Device | Language |

## 2. Psychographics and behavior
- Pain points: [3-5, customer wording]
- Motivation: [gain / avoid]
- Purchase context: [payment, reviews, returns, subscription]
- Online behavior: [platforms, peak hours, preferred formats]

## 3. Targeting settings per platform
### Meta: [age/gender/geo/interest themes/behaviors/exclusions/Advantage+ control]
### Google: [...] · TikTok: [...] · LinkedIn: [...] · Pinterest: [...]

## 4. Audience tiers
| Tier | Audience | Signal | Campaign that uses it |

## 5. Lookalike and seed audiences
| Seed | Size | Lookalike % | Platform | Purpose | Consent status |

## 6. Targeting hypotheses (hand to A/B test)
| # | Hypothesis | Variable | Metric | Priority |

## 7. Sizing and overlap notes
- Estimated reach per ad set: [number]
- Days of spend the pool supports at planned budget: [number]
- Overlaps to avoid: [audiences likely to collide]
- Regional CPM expectation per `references/benchmarks-global.md`: [range]
```

## After the first live data

- After 3-5 days: compare CPA and lead quality per audience segment. Record winners and losers in the profile.
- Winning audiences expand into similar lookalike or broad pools (see `55-scaling-ads-global`). Losing audiences get a written reason so nobody retests the same thing.
- Monthly, reconcile the profile against CRM data: which source closes best. Cheap leads are not always good leads.

## Related skills

- `09-customer-insight-global`: run first — supplies insight, pain, and customer language.
- `08-competitor-research-global`: see who competitors target and with which angles (ad libraries).
- `10-reverse-kpi-global`: max CPA and budget before planning.
- `54-media-plan-global`: turns this audience map into channel and budget allocation.
- `52-account-structure-global`: turns targeting settings into ad set structure.
- `56-retargeting-plan-global`: detailed warm/hot tiering and messaging.

## Quality checklist

- [ ] Profile built from real data (CRM, analytics, surveys), not guesses
- [ ] Market tier stated, with the CPM expectation from `references/benchmarks-global.md`
- [ ] Pain points written in customer language, not marketer language
- [ ] Targeting settings complete and paste-ready for every platform in the plan
- [ ] Three tiers defined (cold/warm/hot) with cross-exclusion rules
- [ ] Lookalike seeds meet minimum size and have documented consent
- [ ] 3-5 targeting hypotheses, each with a test variable and a metric
- [ ] Cold audience large enough to sustain 7 days of planned spend below frequency 2.5
- [ ] Plan in place to update the profile after 3-5 days of live data

---

REFERENCE FILE: references/benchmarks-global.md

# Global Ad Benchmarks 2025–2026

> **Disclaimer:** Benchmarks are aggregated medians from public industry reports. Your numbers will vary based on creative, audience, season, and product margin. Use these as **diagnostic baselines, not targets**.

**Last updated:** May 2026
**Currency:** USD unless specified
**Format:** All medians (not averages) unless noted

---

## Quick Reference — Global Medians (All Industries)

| Metric | Median | Top 25% | Bottom 25% |
|---|---|---|---|
| **CPM (Meta)** | $13–$20 | <$8 | >$25 |
| **CPC (Meta)** | $0.70–$1.72 | <$0.50 | >$2.50 |
| **CTR (Meta)** | 1.57%–2.19% | >3.0% | <0.9% |
| **CVR (Meta)** | 1.57%–9.21% | >10% | <1.5% |
| **ROAS (Meta)** | 1.93–2.79 | >4.0 | <2.0 |
| **CPA (Meta, global)** | $18–$38 | <$15 | >$50 |

*Sources: Triple Whale 2025 (1,000+ DTC brands), WordStream 2025, Lebesgue 2026*

---

## By Region (Meta CPM, 2026)

| Region / Tier | CPM Range | CPC Range | Notes |
|---|---|---|---|
| 🇺🇸 **United States** | $15–$25 | $1.40–$1.85 | Highest globally; $19.66 median 2025 |
| 🇨🇦 **Canada** | $12–$20 | $1.20–$1.60 | Tier 1, slightly below US |
| 🇦🇺 **Australia** | $12–$22 | $1.30–$1.80 | Tier 1, high purchasing power |
| 🇬🇧 **United Kingdom** | $10–$18 | $1.10–$1.50 | Tier 1, Western EU premium |
| 🇩🇪 **Germany / Western EU** | $5–$10 | $0.90–$1.30 | EU median CPM $4.90 in 2025 |
| 🇯🇵 **Japan** | $8–$15 | $1.00–$1.50 | Tier 1, mature market |
| 🇸🇬 **Singapore** | $6–$12 | $0.70–$1.20 | SEA leader |
| 🇹🇭 **Thailand / Malaysia** | $3–$7 | $0.40–$0.80 | SEA emerging |
| 🇻🇳 **Vietnam** | $1–$3 | $0.10–$0.30 | See `benchmarks-vietnam.md` |
| 🇮🇩 **Indonesia / Philippines** | $2–$5 | $0.20–$0.50 | SEA volume markets |
| 🇧🇷 **Brazil / Latin America** | $2–$6 | $0.30–$0.70 | LATAM Tier 2 |

*Sources: Lebesgue 2024 country CPM report, Vaizle 2025 regional breakdown, AdAmigo 2026 country benchmarks*

**Key insight:** Tier 1 (US/AU/CA/Western EU) costs 6–7× more than Tier 2 (SEA/LATAM). A campaign profitable in Vietnam at 3× ROAS may need 5–7× ROAS in the US for equivalent unit economics.

---

## By Industry (Meta, 2026 Lebesgue Data)

| Industry | CPM | CTR | CVR | CPA |
|---|---|---|---|---|
| **Beauty & Health** | $12.46 (highest) | ~2.2% | 7.10% | $25.49 |
| **Clothing & Fashion** | $9–$11 | 2.84% (high) | 3.26% | $21.47 |
| **Art / Home Decor** | $8–$10 | 2.92% (highest) | ~4% | $22 |
| **Food & Beverage** | $8–$10 | 1.85% (lowest) | 2.02% (top) | $30 |
| **Electronics** | $7–$9 | ~2% | 1.20% (lowest) | $45 |
| **Pets & Animals** | $7–$9 | ~2.1% | ~3% | $20 |
| **Sports & Outdoors** | $7–$9 | ~2.2% | ~3.5% | $25 |
| **Home & Garden** | $7–$9 | ~2% | ~4.5% | $28 |
| **Hardware / Automotive** | $6.96 (lowest) | ~1.8% | ~1.5% | $38 |
| **Finance & Insurance** | $10–$15 | ~1.5% | varies (lead) | $50+ (CPL) |
| **B2B / SaaS** | $15–$30 | ~1.0% | varies (lead) | $80–$300 (CPL) |
| **Real Estate** | $8–$12 | 0.99% | 10.68% (lead) | $16.92 (CPL) |

*Sources: Lebesgue 2026, Trendtrack 2025*

---

## ROAS by Industry (Meta, 2025)

| Industry | Median ROAS |
|---|---|
| Automotive | 2.54 (top) |
| Sports & Outdoors | 2.28 |
| Travel & Luggage | 2.25 |
| Home & Garden | 2.18 |
| Beauty & Health | 3.00+ (high CVR offsets cost) |
| Pets & Animals | 1.58 |
| Food & Beverage | 1.56 |
| **Global median** | **1.93–2.79** |

*Source: Triple Whale 2025 DTC dataset*

**Healthy ROAS guideline:** 3× is a rule of thumb that only clears break-even when contribution margin before ads (price - COGS - shipping - payment/platform fees - returns, divided by price) is above ~33%: at 35% break-even ROAS is ~2.9×, at 30% it is ~3.3×, so 3× loses money. For lower margins (commodity goods), aim for 5×+. For high-margin (digital/info products), 1.5× can be profitable.

---

## Seasonal Patterns (Meta Global, 2025)

| Period | Median CPM | vs. Average |
|---|---|---|
| **January** | $15.74 | -21% (post-holiday reset) |
| **Q1 average** | $18.29 | baseline |
| **Q2–Q3** | $18–$21 | rising |
| **October** | $22 | +10% |
| **November (peak)** | $25.22 | +27% |
| **December** | $22.04 | +11% |
| **Q4 average** | $22.98 | +26% above Q1 |

*Source: Superads.ai 2025 dataset, $3B ad spend tracked*

**Strategic implications:**
- 🟢 **Q1 (Jan–Mar):** Cheapest acquisition. Stock budget here if cashflow allows.
- 🔴 **Q4 (Oct–Dec):** Most expensive. Either front-load brand spend in Q3 or reserve 30%+ higher budget.
- 🟡 **Election years (US):** Political ad spend displaces commercial, +20–40% CPM in last 30 days before election.

---

## Other Platforms (2025–2026 Approximate)

| Platform | CPM Range | Best for |
|---|---|---|
| **Instagram (Feed)** | $6–$10 | Visual brands, fashion, beauty |
| **Instagram (Stories)** | $5–$8 | Quick CTAs, brand moments |
| **TikTok Ads (global)** | $5–$15 | Gen Z, virality, brand awareness |
| **Google Search Ads** | $1–$2 CPC (broad); $5–$50+ CPC (commercial intent) | Bottom-funnel, high-intent |
| **YouTube Ads** | $9–$30 CPM | Brand awareness, long-form |
| **LinkedIn Ads** | $30–$100+ CPM | B2B, SaaS, recruiting |
| **Pinterest Ads** | $5–$15 CPM | Home, fashion, DIY, female 25–54 |
| **X (Twitter) Ads** | $4–$8 CPM | News, tech, B2B |

---

## Email Marketing Benchmarks (Global, 2025–2026)

| Metric | Median | Top Performers |
|---|---|---|
| **Open Rate** | 22–27% | >35% |
| **Click-Through Rate (CTR)** | 2.5–3.5% | >5% |
| **Click-to-Open Rate (CTOR)** | 10–14% | >18% |
| **Unsubscribe Rate** | <0.5% | <0.2% |
| **Bounce Rate** | <2% | <1% |
| **Conversion Rate (ecom)** | 1–3% | >5% |

*Sources: Mailchimp 2025 benchmark report, Klaviyo 2025 industry data*

---

## How to Use These Benchmarks

### ✅ DO
- Use as **diagnostic baselines** — "is my CPM 2× median? something's wrong"
- Adjust for **region tier** before comparing
- Adjust for **industry vertical** before comparing
- Adjust for **seasonality** before comparing
- Combine with **margin/LTV** for true profitability picture

### ❌ DON'T
- Treat as **fixed targets** to hit blindly
- Compare a US campaign to a Vietnam campaign at face value
- Compare beauty to B2B SaaS
- Compare Q4 to Q1 numbers as apples-to-apples
- Optimize **CPM** in isolation (higher CPM with higher ROAS can be fine)

---

## Diagnostic Heuristics

When a campaign underperforms, check in this order:

1. **CTR below median?** → Creative problem (hook, visual, format)
2. **CTR good but CVR below median?** → Landing page or offer problem
3. **CVR good but ROAS low?** → AOV/margin problem, not ad problem
4. **CPM 2×+ above region median?** → Audience too narrow, or competitive auction
5. **All metrics good but unprofitable?** → Margin or fulfillment cost problem

---

## Source Notes

Primary references compiled May 2026:

- **Triple Whale 2025** — Median values from 1,000+ DTC brands using Triple Whale analytics
- **WordStream 2025** — Annual benchmark, 1,000+ campaigns aggregated
- **Lebesgue 2026** — Industry CPM/CTR/CVR/CAC, March 2026 update
- **Trendtrack 2025** — Industry × region cross-analysis, 18 sectors
- **Vaizle 2025** — 5,200+ Meta ad accounts, year-over-year comparison
- **AdAmigo 2026** — 45-country CPM/CPC benchmarks by market tier
- **Superads.ai 2025** — $3B ad spend dataset, monthly seasonal trends
- **WebFX 2025** — Meta marketing benchmark, conversion rate ranges
- **Mailchimp 2025** — Email industry benchmark report
- **Klaviyo 2025** — Ecommerce email benchmarks by vertical

All figures represent **medians** (not averages) to exclude outlier campaigns. Where reported, top-25% and bottom-25% percentiles included.

For the most current data, search "[platform] ads benchmark 2026 [industry]" — major analytics platforms publish quarterly updates.

---

**Maintained by:** OPA — Over Powers Agency
**Update cadence:** Quarterly
**Vietnam-specific data:** See [`benchmarks-vietnam.md`](benchmarks-vietnam.md)

---

Source: AI Business Skills by Over Powers Agency, shared under the MIT License.

MIT License

Copyright (c) 2025 Over Powers Agency

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.