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

# Content Audit (Global)

An audit is not a judgment of the past. It is how you replicate winners and retire losers with evidence behind the decision. Every recommendation must attach to a specific action — never a general observation. Run it at least monthly; do not wait for the quarter to close.

## Information gathering

Ask up to 4 questions before starting:

1. **What is the audit period?** Month / quarter / campaign — exact start and end dates.
2. **Which channels?** Instagram / TikTok / LinkedIn / Facebook / X / YouTube / blog / email / all.
3. **Is post-level data available?** Reach, engagement, views, CTR, leads per post — paste or export it. If an analytics MCP or connector is already set up, pull directly.
4. **What is the goal of this audit?** Find winners to scale / find losers to stop / both.

Without post-level data there is no audit. Walk the user through the exports in step 1 first.

## Principles

1. **Audit monthly, not quarterly.** Standard cadence: pull data on the 28th-30th, finish the audit before the 2nd of the following month.
2. **Classify by percentile, not by feel.** Top 20 percent are winners, bottom 20 percent are losers, the middle 60 percent are average.
3. **Compare like with like.** Organic against organic, boosted against boosted, channel against itself. Never pool them.
4. **The primary metric follows the post's objective.** TOFU is measured on reach or views, MOFU on engagement rate or saves, BOFU on clicks or leads. One metric for everything produces nonsense.
5. **Winners become templates.** Document the winning structure — hook type, format, angle, length — so it can be reproduced.
6. **Every finding needs a "why" hypothesis.** "This post did well" is not a finding. "It did well because X" is.
7. **Judge against your own trailing baseline first.** Organic reach and engagement rates vary too much by follower count, niche, and algorithm changes for a published cross-industry number to be a fair target. Use your own trailing 3-month median as the baseline, and use published benchmarks only where they exist for that metric.

## Workflow

### 1. Pull the data (day 28-30)

| Channel | Source | Metrics to export |
|---------|--------|-------------------|
| Instagram | Instagram professional account insights or Meta Business Suite, post level | Reach, accounts engaged, saves, shares, profile visits, link clicks |
| Facebook | Meta Business Suite, post level export | Reach, engagement, link clicks, shares, negative feedback |
| TikTok | TikTok Analytics, Content tab | Views, average watch time, completion rate, shares, follows from video |
| LinkedIn | LinkedIn page or creator analytics | Impressions, engagement rate, click-through, follower gain, reposts |
| YouTube | YouTube Studio | Views, average view duration, impressions click-through rate, subscribers gained |
| X | X analytics | Impressions, engagement rate, link clicks, profile visits |
| Blog / site | GA4 — Explore report by landing page and source/medium | Sessions, engagement rate, average engagement time, key events, assisted conversions |
| Email | Klaviyo, Mailchimp, or equivalent campaign report | Open rate, CTR, click-to-open rate, placed-order rate, unsubscribe rate |
| Leads and revenue | CRM (HubSpot, Pipedrive) or store (Shopify) | Leads and orders attributed to content, by source |

One row per post: publish date, channel, format, pillar, funnel stage, hook type, primary metric, secondary metrics. If the classification columns (pillar, format, hook type) are missing, backfill them by hand from the content calendar (`01-content-calendar-global`).

### 2. Set thresholds and classify

1. Choose the primary metric for each group by objective (TOFU: reach or views; MOFU: engagement rate or saves; BOFU: clicks or leads).
2. Rank posts within each group by that metric.
3. Top 20 percent are WINNERS, bottom 20 percent are LOSERS, the remaining 60 percent are AVERAGE.

Apply the split per channel and per group. A cross-channel ranking measures the channel, not the content.

### 3. Find the pattern across 5 axes

For the winner set and the loser set separately, look for what they share:

| Axis | Question |
|------|----------|
| Pillar | Which pillar do the winners concentrate in? Which pillar produced no winner at all? |
| Format | Video, carousel, static, text, long-form — which format wins on which channel? |
| Hook type | Pain, curiosity, number, contrarian, result-first — which one holds attention? |
| Publish time | Which slots produce winners? |
| Length | Short or long — and does the answer differ by channel? |

### 4. Write the core finding

Two or three sentences answering why the pattern works, checked against the customer insight file (`09-customer-insight-global`). Example of the right shape: "Videos with a number in the hook and a real face on camera hold completion far better than carousels — this audience trusts a person more than a designed asset."

A finding is a hypothesis grounded in data, not an impression.

### 5. Recommend: keep / kill / scale

- **Do now (next week):** replicate the winners — each action with an expected impact.
- **Stop:** the content types, angles, or formats to retire — each with the data reason.
- **Test next:** new hypotheses plus how they will be measured. Hand off to `40-next-content-plan-global` and `19-ab-test-setup-global`.

### 6. Compare KPIs against prior period and benchmark

For each metric record: this period / prior period / baseline / verdict.

Baselines to use:

- Email: open rate median 22-27 percent, top performers above 35 percent; CTR 2.5-3.5 percent; click-to-open 10-14 percent; unsubscribe under 0.5 percent (see `references/benchmarks-global.md`).
- Paid social: use the published CPM, CPC, CTR, and ROAS medians in `references/benchmarks-global.md`, adjusted for region tier, industry, and season before comparing.
- Organic social engagement, reach, and completion rate: no reliable cross-industry global median exists. Use your own trailing 3-month median as the baseline and judge movement against it.

## Output structure

File name: `content-audit-[brand]-[YYYYMMDD].md`

```markdown
# Content Audit — [Period] — [Channels]

## I. Period overview
| Item | Figure |
|------|--------|
| Posts published | |
| Total reach / impressions | |
| Average engagement rate | |
| Leads attributed to content | |
| Revenue attributed to content | |

## II. Classification
### WINNERS — top 20%
| Post | Channel | Format | Pillar | Hook type | Standout metric | Why it worked (hypothesis) |
|------|---------|--------|--------|-----------|-----------------|----------------------------|
| | | | | | | |

### LOSERS — bottom 20%
| Post | Channel | Format | Pillar | Hook type | Weak metric | Why it failed (hypothesis) |
|------|---------|--------|--------|-----------|-------------|----------------------------|
| | | | | | | |

### AVERAGE — middle 60%
[2-3 sentences of general observation]

## III. Pattern analysis
**Winner pattern:** [recurring pillar / format / hook / time / length]
**Loser pattern:** [what the failures share]
**Core finding:** [2-3 sentences — why it works, checked against customer insight]

## IV. Recommendations
**Do now (next week):**
1. [Action + expected impact]

**Stop:**
1. [What to retire — data reason]

**Test next:**
1. [Hypothesis + how it will be measured]

## V. KPI comparison
| Metric | This period | Prior period | Baseline | Verdict |
|--------|-------------|--------------|----------|---------|
| Instagram reach per post | | | trailing 3-month median | |
| TikTok average completion rate | | | trailing 3-month median | |
| LinkedIn engagement rate | | | trailing 3-month median | |
| Email open rate | | | 22-27% median | |
| Email CTR | | | 2.5-3.5% median | |
| Leads from content | | | | |
```

## Related skills

- `13-data-analysis-global`: pulling and cleaning the data before the audit, including connector and MCP setup.
- `03-performance-eval-global`: paid media evaluation — this audit covers organic.
- `07-marketing-report-global`: the audit result is an input to the monthly report.
- `40-next-content-plan-global`: run immediately after. An audit that does not produce a new plan was wasted effort.
- `01-content-calendar-global`: winners return to the calendar as base templates.
- `09-customer-insight-global`: check patterns against insight to explain why something worked.

## Quality checklist

- Post-level data present — no auditing on impressions of the whole period.
- Primary metric chosen per objective group (TOFU / MOFU / BOFU judged differently).
- Percentile classification applied: top 20 / bottom 20 / middle 60, within channel and group.
- Like compared with like — organic separated from boosted.
- Pattern analysis covers all 5 axes: pillar, format, hook, publish time, length.
- Every winner and loser carries a "why" hypothesis checked against insight.
- Recommendations split into three groups: do now / stop / test next.
- Every action has an expected impact or a measurement method.
- KPI comparison includes prior period, plus a published benchmark where one exists and a trailing self-baseline where one does not.
- Winner structures documented as reusable templates.
- `40-next-content-plan-global` scheduled to run immediately after.

---

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.