Ad Performance Analyser for AI Coding Agents

What Is the Ad Performance Analyser?Link to this section

The Ad Performance Analyser is a free, open-source agent skill that turns the ad data you already have into a decision. Hand your AI agent a CSV export, pasted numbers or a screenshot from Google, Meta, LinkedIn or TikTok. It checks the data, tells you whether the spend is coming back and where the next dollar should go.

It is one of six skills in the Flyweel skills repository, released under Apache 2.0. Unlike the other five, it needs no Flyweel account and no MCP connection.

How Do I Install It?Link to this section

Install just this skill:

npx skills add flyweel-production/flyweel-skills --skill ad-performance-analyser-flyweel

Or install all six Flyweel skills:

npx skills add flyweel-production/flyweel-skills

It works with Claude Code, Cursor, Codex, Gemini CLI, OpenCode and most other coding agents. In Claude Code you can also install the plugin with /plugin marketplace add flyweel-production/flyweel-skills and /plugin install flyweel@flyweel.

Restart your agent after installing so it picks up the skill.

What Does It Check Before Analysing?Link to this section

Before any analysis, the skill states what it was given:

• Period: the dates covered, and whether the latest week is complete.
• Currency: mixed currencies stay apart. If the file names none, it says so.
• What “conversions” counts: if the export does not say, it states its assumption and lists it as a question for you.
• Revenue: if deal or revenue data is missing, cost per result is the deepest answer available. Missing revenue is never treated as zero.

It also checks sample size. Under 100 clicks, it draws no conclusion from click rate. Under 30 results, cost per result is a rough guide. Under 10, it will not rank on it. Under 7 days is too short for a trend.

What Questions Can It Answer?Link to this section

• “Are my ads working?” Compares the last 30 days with the 30 before and with your target cost per result.
• “Cost per lead went up. Why?” Splits by campaign, finds the week it moved, then checks whether impression cost, click rate or conversion rate drove it.
• “Where should I spend more or less?” Looks at what each extra result cost when spend rose, not the average.
• “How do I split budget between channels?” Compares the same result in the same currency, and cost per won deal where deal data exists.
• “Is this campaign wearing out?” Checks frequency, falling click rate and rising cost per click while impression cost holds.
• “What is my real return?” Gives up to three figures, most trustworthy first: revenue from ad-sourced deals, blended return, then the platform’s own claim.
• “Plan next month’s budget.” Models flat, plus 20 percent and minus 20 percent from 90 days of weekly data, and gives a range.

It can also write a weekly summary or monthly review from the same data.

How Deep Does It Go?Link to this section

The skill matches depth to the question:

• Health check: spend against plan, cost per result against target and the recent trend. Answer: on track or off, and the one lever that matters most.
• Review: adds a channel breakdown, the top and bottom campaigns by cost per result, a four-week trend and three to five actions.
• Deep dive: adds the funnel from impression to revenue, results by launch date or audience, and where the next block of budget should go.

Which Red Flags Does It Raise?Link to this section

The skill raises these whenever they appear, asked or not:

• A campaign taking more than 20 percent of spend with no results.
• A campaign whose cost per result is more than twice the account average, or up more than 30 percent week on week.
• Platform-reported results far above what your CRM or sales records show.
• One campaign taking more than 60 percent of spend.
• Frequency above 5 on Meta.
• Impression share under 50 percent on a Google brand campaign.

What Benchmarks Does It Use?Link to this section

Your account’s own history comes first. The skill uses these rough ranges only when there is no history to compare against, and says when it does:

Platform	Click rate	Cost per click	Cost per 1,000 impressions
Google Search	3-6%	$1-5 (legal, finance and insurance $5-50)	n/a
Google Display / Performance Max	0.3-1%	$0.30-2	$2-10
Meta (Facebook / Instagram)	0.8-2%	$0.50-3	$5-15
LinkedIn	0.4-0.8%	$5-15	$30-80
TikTok	0.5-1.5%	$0.30-2	$3-10

Vertical, country, funnel stage and creative move these a long way.

What Does the Answer Look Like?Link to this section

The finding comes first, then the numbers. Comparisons go in a table, with the period and currency stated every time and change shown as a signed percentage. Each answer ends with three or four numbered actions. If the data cannot answer the question, the skill says what is missing.

Example PromptsLink to this section

• “Here is my campaign export. Why did cost per lead go up?”
• “Here’s my Google Ads data from last month. Give me a health check.”
• “Compare my LinkedIn and Google Ads cost per lead. Which channel should I scale?”
• “Is this Meta campaign wearing out?” Attach a screenshot or paste the weekly numbers.
• “Plan next month’s budget from these 90 days.”
• “Write a monthly review from this export.”

---

Want live data instead of exports? MCP Deep Ad Analysis reads Google, Meta and TikTok through the Flyweel MCP server. See all six skills in the Flyweel agent skills docs or read the launch announcement.

Ad Performance Analyser FAQs

Common questions about the free agent skill for reviewing ad data from Google, Meta, LinkedIn and TikTok.

What is the Ad Performance Analyser?

The Ad Performance Analyser is a free, open-source agent skill that reviews ad data you hand to your AI agent: a CSV export, pasted numbers, a screenshot or an API reply from Google Ads, Meta Ads, LinkedIn Ads or TikTok Ads. It checks the data first, then answers whether the ads are working, why cost per lead moved and where to spend more or less. It ends every analysis with specific actions.

What is a SKILL.md?

A SKILL.md is a markdown file of instructions an AI agent loads when your question matches it. It tells the agent what to check, which figures to trust and how to answer. The Flyweel skills are open source, so you can read exactly what your agent will do before you install them.

How do I install the Ad Performance Analyser?

Run npx skills add flyweel-production/flyweel-skills --skill ad-performance-analyser-flyweel in your terminal, then restart your agent. It works with Claude Code, Cursor, Codex, Gemini CLI, OpenCode and most other coding agents. Leave off the --skill flag to install all six Flyweel skills.

What ad platforms does it support?

Google Ads (Search, Display and Performance Max), Meta Ads, LinkedIn Ads and TikTok Ads. It compares each account against its own history first. It only falls back to rough platform ranges, such as a 3 to 6 percent click rate on Google Search, when the account has no history, and it says when it does.

Do I need a Flyweel account?

No. The Ad Performance Analyser works from the data you supply, with or without a Flyweel account or MCP connection. It is free and open source.

Is it open source?

Yes. It ships in the flyweel-production/flyweel-skills repository on GitHub under the Apache 2.0 licence, alongside five other Flyweel skills.

How is this different from the MCP Deep Ad Analysis skill?

The Ad Performance Analyser works from data you paste or upload. MCP Deep Ad Analysis reads live Google, Meta and TikTok data through the Flyweel MCP server, so it needs a Flyweel account and a connected ad platform. Use this skill for an export you already have; use the MCP skill for live, connected data.