Creative Analytics

A woman named Dana manages Meta ads for a DTC hair care brand in Charlotte. She’s been paying $250 a month for Motion, the creative analytics platform that every performance marketer on Twitter swears by. Motion is good. It tags her creatives automatically, surfaces her top performers, flags creative fatigue before it tanks her CPA, and generates visual reports she can send to her clients.

But Dana has a problem. She’s a one-person operation running three client accounts. That’s $250 per workspace, per month. $750/mo for creative analytics. $9,000 a year.

Last month Dana connected her Meta Ads accounts to Claude Code using a free MCP server, wrote three skill files, and built herself a creative analytics system that does 80% of what Motion does for $0/mo in tooling costs. The other 20% (Motion’s polished visual dashboards and AI-generated video frame analysis) she doesn’t need. What she does need is: which ads are winning, which are dying, what patterns explain the difference, and what to make next. Claude gives her all four.

This is the playbook for building Dana’s system. It uses Claude Code, a Meta Ads MCP connector (several free options exist), and three skill files that turn raw performance data into kill/scale/iterate decisions.

---

What Motion Actually Does (So You Know What You’re Replacing)

Before we build the replacement, let’s be honest about what Motion offers and where the DIY version matches it, beats it, or falls short.

What Motion does well:

Motion connects to your Meta, TikTok, YouTube, and LinkedIn ad accounts. It pulls every creative asset, automatically tags each one by format type (static, video, carousel), hook style (question, stat, testimonial, claim), visual elements (product shot, lifestyle, UGC), and several other attributes. Then it overlays performance data (CTR, CPC, CPA, ROAS, spend) and lets you group, filter, and compare by any combination of tags.

The output: you can see that “UGC-style videos with question hooks are outperforming studio product shots by 2.3x on ROAS” without manually tagging a single ad. Motion’s AI does the tagging. You just read the dashboard.

Motion also tracks creative fatigue. When an ad’s CTR starts declining week over week while frequency rises, Motion flags it as fatiguing before your CPA spikes. That early warning is worth real money.

What Motion costs:

As of April 2026, Motion’s Starter plan is $49/mo (limited to basic reports on one ad account). The plan most people actually need starts at $250/mo. For agencies with multiple client accounts, you’re paying per workspace. The math gets ugly fast.

Where Claude matches or beats Motion:

Performance analysis, pattern recognition, and creative strategy recommendations. Claude can read your live Meta Ads data, identify winners and losers by any metric, parse your naming conventions to group results by hook/persona/style, detect fatigue patterns, and generate specific recommendations for your next batch of creatives. And it does all of this through natural language. You ask a question, you get an answer.

Where Motion still wins:

Visual dashboards. Motion’s reports look professional and are designed to be shared with clients or stakeholders. Claude gives you text-based analysis and can generate charts as artifacts, but it’s not a presentation tool. If you need polished visual reports for client calls, Motion is still worth the money.

Frame-by-frame video analysis. Motion can analyze specific moments in video ads (which frame loses viewers, which hook retains attention). Claude can’t watch your videos. This matters if video is your primary format. Less so if you’re running mostly static ads (which is what Pillar 1 produces).

Bottom line: If you run static ads and manage your own accounts (or a small number of client accounts), Claude replaces Motion. If you’re an agency running video-heavy campaigns across 20 clients who need polished visual reports, keep Motion. Know what you’re buying.

---

What You Need

1. Claude Code with Pro or Max subscription
2. A Meta Ads MCP connection (three free options, explained below)
3. The naming convention from Pillar 2 (if you followed it, this step is automatic)
4. Your ad account running ads (you need at least 7-14 days of performance data for the analysis to be meaningful)

---

Step 1: Connect Claude Code to Your Live Meta Ads Data

You have three good free options here. Pick one.

Option A: Pipeboard (Recommended for Most People)

Pipeboard is a Meta Business Partner with a free tier that covers everything in this guide. It handles OAuth, rate limiting, and pagination. You never touch Meta’s API directly.

In your terminal:

claude mcp add --transport http pipeboard-meta-ads https://meta-ads.mcp.pipeboard.co/

Restart Claude Code. When you run your first query, Pipeboard will redirect you to Facebook login to authorize access. Select your ad account, approve the permissions, and you’re connected.

What you get: 60+ tools for campaign management, creative analysis, audience insights, and reporting. Read and write access (you can pull data and also make changes like pausing ads, but changes always require confirmation).

Option B: Windsor MCP

Windsor focuses on multi-touch attribution and cross-channel mapping. It connects 50+ data sources and gives Claude access to 710+ data fields across account, campaign, ad set, ad, and creative levels.

Go to windsor.ai, create an account, connect your Meta Ads account. Then add the Windsor MCP connector through Claude’s settings (Settings > Integrations > Add Connector, search for Windsor).

Best for: Advertisers who also run Google Ads and want cross-channel attribution analysis alongside creative analytics.

Option C: Composio (If You Already Set It Up in Pillar 2)

If you followed Pillar 2, you already have Composio’s Metaads MCP connected to Claude Code. You don’t need to do anything else. The same connection that handles bulk uploading also provides read access to all your performance data.

Best for: Anyone who already has Composio running and doesn’t want to add another tool.

Verify Your Connection

Regardless of which option you chose, test it:

Pull my Meta Ads performance summary for the last 7 days. 
Show total spend, impressions, clicks, CTR, CPC, and ROAS at the 
account level.

What you should see: Claude returns a clean summary table with your account’s last 7 days of data. If you see numbers, you’re connected.

If you see an error: Most common cause is that the OAuth token expired or the ad account wasn’t selected during authorization. Reconnect through the MCP provider’s dashboard.

---

Step 2: Create the Creative Analyzer Skill

This is the core skill file that turns Claude into a creative analytics engine. It defines how Claude should analyze your ads, what patterns to look for, and how to present results.

Create a Claude Code skill file called "creative-analyzer" with these 
instructions:

PURPOSE:
Analyze Meta Ads creative performance data to identify winners, losers, 
fatigue patterns, and actionable insights. Replaces Motion app's core 
creative analytics functionality.

ANALYSIS FRAMEWORK:

1. WINNER IDENTIFICATION
   - Winners: ads with ROAS above account average AND spend > $50
   - Stars: ads with ROAS in top 10% AND spend > $200
   - Consistent performers: ads maintaining above-average ROAS for 14+ days

2. LOSER IDENTIFICATION  
   - Below-average ROAS with spend > $100 (enough data to be conclusive)
   - Flag any ad spending > $50/day with ROAS below 0.5 as "urgent kill"

3. FATIGUE DETECTION
   - Flag when: CTR declined 20%+ vs. first 7 days of delivery
   - Flag when: frequency exceeds 3.0 AND CTR is declining
   - Flag when: CPA increased 25%+ vs. first 7 days
   - Status labels: HEALTHY, WARNING, CRITICAL

4. PATTERN ANALYSIS (requires naming conventions)
   - Parse ad names using pattern: [brand]_[objective]_[hook]_[persona]_[style]_[version]_[date]
   - Group performance by: hook type, persona, visual style
   - Identify: best-performing hook, best-performing persona, best-performing style
   - Identify: worst-performing combinations (where to stop spending)

5. RECOMMENDATIONS
   - For each CRITICAL fatigue ad: recommend specific refresh action
   - For each winner: recommend 2-3 variations to test
   - For overall strategy: recommend next week's creative brief based on patterns

OUTPUT FORMAT:
Always structure the analysis as:
A. Executive summary (3 sentences: what's working, what's dying, what to do)
B. Winners table (ad name, spend, ROAS, CTR, days active, status)
C. Kill list (ads to pause immediately, with reason)
D. Fatigue alerts (ads approaching fatigue, with recommended action)
E. Pattern insights (which hooks/personas/styles are winning)
F. Next steps (specific creative brief for next batch)

Save to .claude/skills/creative-analyzer.md

---

Step 3: Run Your First Creative Analysis

With the MCP connected and the skill file in place, you’re ready to run your first analysis. This is the prompt that replaces your Motion dashboard check:

Run a full creative analysis on my Meta Ads account for the last 14 days.

Pull ad-level data including: ad name, ad set name, campaign name, spend, 
impressions, reach, frequency, clicks, CTR, CPC, conversions, CPA, 
purchase value, ROAS, and days active.

Use the creative-analyzer skill to:
1. Identify winners and stars
2. Flag any urgent kills
3. Detect creative fatigue
4. Parse naming conventions to find hook/persona/style patterns
5. Generate a creative brief for next week based on what's working

Show me the full analysis.

What you should see: Claude pulls your live data, runs the analysis framework, and returns a structured report. The executive summary tells you what’s working and what’s dying in three sentences. The winners table shows your top performers. The kill list tells you what to pause. The pattern analysis tells you which hooks and personas to double down on. And the creative brief tells you exactly what to make next.

First-time calibration: Your first analysis will need some tuning. If Claude’s thresholds don’t match your account (maybe your average ROAS is 1.2, not 2.0), update the skill file:

Update my creative-analyzer skill: my target ROAS is [X]. Redefine 
"winner" as any ad with ROAS above [X] with spend over $[Y]. Redefine 
"urgent kill" as any ad with ROAS below [Z] spending more than $[W]/day.

The skill file learns your specific account standards.

---

Step 4: Set Up the Weekly Report

Motion sends you weekly emails with your creative leaderboard. Here’s how to build the same habit loop in Claude Code:

Create a Claude Code skill called "weekly-creative-report" that does 
the following:

1. Pull the last 7 days of ad-level performance data
2. Compare this week vs. last week for every ad (spend, CTR, CPA, ROAS)
3. Generate a "Creative Leaderboard" showing top 10 ads ranked by ROAS 
   with week-over-week trend arrows (scaling up, declining, stable, new)
4. Generate a "Fatigue Watch" section listing any ad that moved from 
   HEALTHY to WARNING or WARNING to CRITICAL this week
5. Generate a "Kill Confirmation" section listing ads I should pause 
   with the specific reason
6. Generate a "Winners to Scale" section listing ads that are performing 
   well and could take more budget, with a suggested new daily budget
7. End with a 3-sentence executive summary

Save to .claude/skills/weekly-creative-report.md

Every Monday morning, open Claude Code and type:

/weekly-creative-report

Two minutes later, you have your entire creative analytics review for the week. Compare that to logging into Motion, waiting for the dashboard to load, clicking through five different reports, and exporting a PDF. Same information, different delivery mechanism.

---

Step 5: Build the Creative Strategy Recommender

This is where Claude goes beyond what Motion does. Motion tells you what’s working. This skill tells you what to make next and why.

Create a Claude Code skill called "creative-strategist" that takes 
the output of creative-analyzer and generates:

1. A HOOK SCORECARD showing average ROAS, CTR, and CPA for each hook 
   type in my naming convention, ranked best to worst

2. A PERSONA SCORECARD showing the same metrics grouped by persona

3. A STYLE SCORECARD showing the same for visual style

4. A COMBINATION MATRIX showing every hook x persona x style combination 
   that has been tested, with performance data for each. Highlight 
   untested combinations as "gaps to test."

5. A CREATIVE BRIEF for the next batch of ads that includes:
   - The top 3 performing combinations to create more variations of
   - The top 3 untested combinations to experiment with
   - For each combination: a specific headline direction and visual 
     description, ready to feed into the Static Ads Engine (Pillar 1)

Save to .claude/skills/creative-strategist.md

Run it:

/creative-strategist

What you should see: A complete strategic analysis that tells you not just what worked, but what to make next. The combination matrix is especially powerful. If you tested 32 combinations and 12 haven’t been tried yet, Claude identifies those gaps and suggests which ones are most likely to win based on the patterns in your existing data.

This is the feedback loop that compounds. Pillar 1 generates the ads. Pillar 2 uploads them. Pillar 3 analyzes results. And the output of Pillar 3 feeds directly back into Pillar 1 as the creative brief for next month’s batch.

---

Step 6: The Naming Convention Payoff

If you set up the naming convention from Pillar 2 ([brand]_[objective]_[hook]_[persona]_[style]_[version]_[date]), everything in this pillar works automatically. Claude parses your ad names into their component parts and groups performance data by each dimension without you manually tagging anything.

If you didn’t set up naming conventions (or your existing ads have random names), you can still use this system. You’ll just need to do the first analysis manually:

Here are my currently running ads. For each one, I'll tell you the hook 
type, persona, and style. Tag them in your analysis:

- "Summer Glow Campaign v2" = benefit hook, persona: sarah, style: lifestyle
- "Before After Results" = social proof hook, persona: dana, style: minimal
[continue for each ad]

Use these tags for the creative-analyzer and creative-strategist skills.

This works, but it’s manual and tedious. The naming convention exists so you never have to do this. If you’re starting fresh, set it up from day one. If you have legacy campaigns, rename your ads to follow the pattern before your next analysis run. Ten minutes of renaming saves hours of manual tagging over the next year.

---

When to Use This vs. When to Keep Motion

Use Claude for creative analytics when:

You run mostly static ads. You manage 1-5 ad accounts. You don’t need polished visual dashboards for stakeholder presentations. You want strategic recommendations, not just data. You’re budget-conscious and $250/mo per workspace adds up.

Keep Motion when:

You run video-heavy campaigns and need frame-by-frame retention analysis. You manage 10+ client accounts and need visual reports you can share via URL. Your clients or stakeholders expect polished dashboards (not text-based analysis). You need TikTok, YouTube, and LinkedIn creative analytics in the same workspace. You value Motion’s AI auto-tagging across creative attributes beyond naming conventions.

Use both when:

You use Motion for its visual reports and video analysis on your biggest accounts, and Claude for the strategic layer (what to make next, which combinations to test, where the gaps are) that Motion doesn’t provide.

---

The SaaS Alternatives (For People Who Don’t Want to Build)

If you want the “replace Motion” outcome but don’t want to build skill files, here are the ready-made alternatives:

Uplifted (Free tier available, Pro from $49/mo): AI-powered creative tagging, asset-level analytics, and visual reports. The closest direct Motion competitor with a free tier. Works with Meta. Worth testing if you want a dashboard and don’t want to build in Claude.

AdSkate ($129/mo): Pre-campaign creative testing (analyze before you launch, not just after). Google Ads integration that Motion lacks. Strong for teams who want to predict performance before spending.

Superads (Free tier): Multi-account dashboard with custom breakdowns, date comparisons, and flexible grouping. Good for agencies managing multiple accounts who need a visual layer.

None of these give you Claude’s strategic recommendation engine. They’re dashboards. Claude is a strategist. The ideal setup for most serious advertisers is a cheap dashboard tool (Uplifted or Superads at free/low cost) for visual review, plus Claude for the strategic thinking.

---

The Downloadable Artifacts

This pillar comes with three skill files:

1. creative-analyzer.md: The core analysis framework. Drop into .claude/skills/ and run /creative-analyzer to get a full creative performance report on any time range.
2. weekly-creative-report.md: The Monday morning habit. Generates a leaderboard, fatigue alerts, kill list, and executive summary.
3. creative-strategist.md: The recommendation engine. Outputs scorecards, a combination matrix, gap analysis, and a ready-to-use creative brief for your next batch.

All three are in the Vault downloads section.

---

Quick Reference: Cost Comparison

Solution	Monthly cost	What you get
Motion Starter	$49/mo	Basic reports, one account, limited features
Motion Pro	$250+/mo	Full analytics, AI tagging, visual dashboards
Claude + Pipeboard	$0 (MCP is free)	Full analysis, strategy recs, unlimited accounts
Claude + Windsor	$0 (free tier)	Same as above, plus cross-channel attribution
Uplifted	Free-$49/mo	Visual dashboard + AI tagging, no strategy layer
AdSkate	$129/mo	Pre-campaign testing + post-launch analytics

---

What’s Next

You now have the full production-to-analysis loop:

Pillar 1 generates the ads. Pillar 2 uploads them in bulk. Pillar 3 tells you what’s working, what’s dying, and what to make next. That loop alone puts you ahead of 90% of Meta advertisers.

The next pillars complete the operating system:

Pillar 4: Automated Advertorials builds the landing page your static ads point to. One Claude Code session turns any winning competitor advertorial into a finished HTML page for your brand. Because a great ad pointing to a bad landing page is money on fire.

Pillar 5: The Impressions Spy shows you which competitor ads are actually scaling, so you can reverse-engineer winning creative angles before you test them yourself. It’s the research step that goes before the production step.

[CTA: Read Pillar 4: Automated Advertorials] | [CTA: Browse the Vault]