13 Data Analysis Global

What the source does

• FilesystemNone observedNo reads or writes in the source
• NetworkNone observedNo outbound calls in the source
• ShellNone observedNo commands run in the source
• DependenciesPinnedNothing floats
• SecretsNone foundNo keys in source

Five audit layers

Gates applied: no behavioural pass.

Audit 0.4.1 · source sha d435c7de76d1 · 15 Oct 2026Report an issue or request a re-scan

• Publisherminhnv0807 GitHub userOwner of the repository this skill is read from
• Source folderminhnv0807/ai-business-skills/skills/en/13-data-analysis-globalThe exact folder we audited
• LicenceMITAs the repository states it
• Commit pinned for this verdictd435c7de76d1A new audit runs when the source changes
• Repository statusNot archivedLast push 32 days ago
• Sandboxed runNot runSkills are read, not executed; layer L4 is skipped

Its description: Use when raw data exists — Meta, Google, TikTok, GA4, Shopify, a CRM export, or a spreadsheet — and has to become insight and decisions: descriptive, diagnostic, predictive, and prescriptive layers, cuts by channel,…

---
name: 13-data-analysis-global
description: "Use when raw data exists — Meta, Google, TikTok, GA4, Shopify, a CRM export, or a spreadsheet — and has to become insight and decisions: descriptive, diagnostic, predictive, and prescriptive layers, cuts by channel, campaign, creative, audience, and time, cohorts, and a decision log. Trigger on 'analyze this data', 'read these numbers for me', 'what does this export say', 'pull insight from GA4', 'cohort analysis', 'here is the spreadsheet'. Also use when the user pastes a table and asks what it means. Not for — diagnosing ad root cause, see `03-performance-eval-global`; writing the report a stakeholder reads, see `07-marketing-report-global`; auditing account setup, see `21-ads-audit-global`."
metadata:
  version: 2.5.1
  category: performance
  language: en
triggers:
  - "data analysis"
  - "analyze data"
  - "marketing analytics"
  - "Meta Ads analysis"
  - "TikTok Ads analysis"
  - "GA4 report"
  - "performance analysis"
  - "Triple Whale"
  - "Hyros"
  - "Northbeam"
output: A .md report structured as Descriptive, Diagnostic, Predictive, Prescriptive — with tables and concrete recommendations
related:
  - 03-performance-review-global
  - 07-marketing-report-global
  - 10-reverse-kpi-calc-global
  - 12-landing-page-brief-global
---

# Marketing Data Analysis (Global)

> Insight before numbers. Lead with judgment, illustrate with data — never list numbers without interpretation.

---

## Information Gathering

Ask up to 4 questions:

1. **Data source?** Meta Ads, TikTok Ads, GA4, Shopify, Triple Whale/Hyros/Northbeam, Google Sheets — single source or combined?
2. **Time window?** This week, this month, A vs B (e.g. March vs April)?
3. **Current business goal?** Increase leads, lower CPL, raise ROAS, or a specific issue to fix?
4. **Paste data here** — drop a table, or describe core metrics (spend, impressions, clicks, leads, revenue).

---

## Analysis Principles

### Reading Order

```
1. DESCRIPTIVE   — What happened? (numbers, trends)
2. DIAGNOSTIC    — Why? (root cause)
3. PREDICTIVE    — What's next? (forecast)
4. PRESCRIPTIVE  — What to do? (concrete actions)
```

### Presentation Rules

| Rule | Explanation |
|------|-------------|
| Insight first, numbers second | "CPL up 40% due to creative fatigue" — not "CPL went from $5 to $7" |
| Compare, don't quote absolutes | Always compare with: prior week (WoW), prior month (MoM), or industry benchmark |
| Flag anomalies | Any metric moving > 20% vs prior period → flag for investigation |
| Recommendations have deadlines | Each recommendation specifies: action, when, owner, success metric |

---

## Analysis Frameworks by Source

### Meta Ads

| Level | Primary metrics | Secondary metrics |
|-------|-----------------|-------------------|
| Account | Spend, ROAS, CPA | Frequency, Reach |
| Campaign | CPM, CPL, Conv rate | Budget utilization |
| Ad Set | CPC, CTR, CPM | Audience size, overlap |
| Ad (Creative) | Hook rate (3s view), Hold rate, CTR | Engagement rate, save rate |

**Reading Meta Ads:**

```
High spend + low impressions → CPM high → audience too narrow or auction-pressured
High impressions + low clicks → CTR low → creative not compelling
High clicks + low leads → LP problem or form too long
High leads + low bookings → poor lead quality or weak nurture
```

### TikTok Ads

| Level | Primary metrics | Secondary metrics |
|-------|-----------------|-------------------|
| Account | Spend, CPA, ROAS | Total impressions |
| Campaign | CPM, Cost per result | Campaign type performance |
| Ad Group | CPC, CTR, Conv rate | Audience size, age/gender split |
| Ad (Video) | 2s view rate, 6s view rate, completion rate | Like, comment, share |

**Reading TikTok Ads:**

```
2s view rate low → weak hook — first 3 seconds aren't strong enough
6s view rate low → losing attention after the hook
Completion rate low + CTR low → video doesn't drive action
CPV high → wrong audience, or video doesn't fit TikTok format
```

### Google Analytics 4

| Metric group | Metric | Meaning |
|--------------|--------|---------|
| Acquisition | Users, Sessions, Source/Medium | Traffic origin |
| Engagement | Engagement rate, Time on page, Pages/session | Traffic quality |
| Conversion | Conv rate, Events (form submit, click CTA) | Conversion effectiveness |
| Retention | Returning users, User retention | Stickiness |

**Reading GA4:**

```
Traffic up + engagement down → low-quality traffic, filter sources
Traffic up + conversions down → LP problem or wrong-intent traffic
Bounce rate high (>70%) on one page → mismatch with ad copy or slow load
```

### E-commerce Attribution Tools (Dropshipping/DTC)

For dropshipping or DTC stores, native ad-platform metrics often diverge from real revenue. Use one of these:

| Tool | Best for | Key feature |
|------|----------|-------------|
| **Triple Whale** | Shopify DTC | Pixel-based attribution, blended ROAS, AI insights |
| **Hyros** | Info products + DTC | Server-side tracking, long-window attribution |
| **Northbeam** | High-spend DTC ($100K+/mo) | MTA + MMM, incrementality testing |
| **Polar Analytics** | Mid-market DTC | All-in-one dashboards, source-of-truth tracking |
| **Wicked Reports** | Email-heavy DTC | Multi-touch attribution including email |

**Cross-checking:** when Meta reports 5x ROAS but Shopify reports 2x ROAS, trust the platform-of-record (Shopify). The gap is usually iOS 14+ attribution loss.

### Spreadsheet Data (Manual)

When user pastes data from a sheet:

1. Identify core columns: date, channel, spend, units (impressions/clicks/leads/orders), revenue
2. Compute derived metrics: CPL, CPA, ROAS, conversion rate
3. Sort by time to surface trends
4. Group by channel/campaign for comparison

---

## Trend Detection

### Week over Week (WoW)

| Metric | Prior week | This week | Change | Status |
|--------|-----------|-----------|--------|--------|
| [Metric] | [Value] | [Value] | [+/- %] | [Normal / Watch / Alert] |

**Alert thresholds:**
- 10–20% chan

Excerpt, first 6,000 characters of SKILL.md, verbatim from the audited commit. It is text from someone else’s repository, quoted so you can judge it before your agent reads it.

git clone https://github.com/minhnv0807/ai-business-skills
git -C ai-business-skills checkout d435c7de76d1

Works in

1. 2026-10-15
   ✓SAFE B 89 / 100

   first audit · source d435c7de76d1

Questions about 13 Data Analysis Global

• github.com/minhnv0807/ai-business-skills/tree/master/skills/en/1 — the audited source at commit d435c7de76d1 and every finding aboveObserved · 15 Oct 2026

Spotted something wrong? Report a correction · Methodology