Step-by-step guide

Turn Claude into your performance analyst

Connect your Meta Ads and Google Ads accounts to Claude through Detrics. In 5 minutes, you'll have an analyst reviewing your campaigns 24/7.

Meta Ads + Google Ads → Claude

Full setup: ~5 minutes

This guide shows you how to connect your ad platforms to Claude using the Detrics MCP. When you're done, you'll be able to ask Claude things like "how did my Meta campaigns perform this week vs last week?" and get analysis based on real data, not hallucinations.

1

Connect your data sources Meta Ads + Google Ads

Go to the Data Sources section in Detrics and connect your Meta Ads and Google Ads accounts. If you already had them connected, verify they're active.

app.detrics.io/datasources
Meta Ads Meta Ads
Google Ads Google Ads
Data Sources in Detrics showing Meta Ads and Google Ads connected
Data Sources section in Detrics: connect Meta Ads and Google Ads
  • Click on the data source you want to connect (Meta Ads, Google Ads)
  • Authorize access with your Google / Meta account
  • Verify the status shows Connected (green)
✓ If you're already a Detrics user with connected accounts, you can skip straight to Step 2.
2

Create an AI Context Group your client's accounts

A Context is a grouping of accounts that belong to the same client or project. For example: the Meta Ads + Google Ads accounts for "Client X" form one context. This way, when you ask Claude about that context, it knows exactly which data to query.

app.detrics.io/mcp/contexts
AI Contexts view in Detrics with sidebar and New Context button
AI Contexts view in Detrics: each context groups a client's accounts. Click "+ New Context" to create one.
  • Click on "Create Context"
  • Give it a descriptive name (e.g., "My company" or "Client ABC")
  • Select the Meta Ads and Google Ads accounts that belong together
  • Save the context
★ You can create multiple contexts — one per client. Ideal if you manage several accounts as an agency or freelancer.
3

Set up the Detrics MCP in Claude ~30 seconds, no installation needed

Now we connect Claude with Detrics. The MCP (Model Context Protocol) gives Claude direct access to your marketing data. You don't need to install anything — it's configured from claude.ai.

  1. Open claude.ai and log in with your account
  2. Click on your profile (bottom left) → Settings
    Claude - click on Settings from the profile menu
    Open Settings from your profile menu
  3. In the sidebar, click on Connectors
    Connectors page in Claude Settings
    Connectors page in Claude Settings
  4. Click on "Add custom connector"
    Add custom connector button on the Connectors page
    Click "Add custom connector" at the bottom of the list
  5. Enter the connector details:
    Name
    Detrics
    URL
    https://mcp.detrics.io
    Add custom connector dialog in Claude with Name: Detrics and URL: https://mcp.detrics.io
    Enter "Detrics" as the name and "https://mcp.detrics.io" as the URL
  6. Click Connect
    Detrics appears as a custom connector - click Connect
    Detrics appears as a custom connector — click "Connect"
  7. Authorize access — follow the Detrics authorization flow
    Detrics MCP authorization screen
    Authorize Claude to access your Detrics data
  8. Verify it works: open a new chat and ask Claude What contexts do you see for Detrics?
    Connectors in Claude with Detrics enabled
    Detrics appears as an active connector (blue toggle) in Claude
✓ You don't need to install Claude Desktop. Everything works from claude.ai in the browser.
4

Create your Performance Review project The step that makes the magic happen

Now we'll create a Project in Claude with specific instructions so it behaves like an expert performance marketing analyst. This is what transforms Claude from a generic chatbot into your personal analyst.

How to create the project

  1. In the left sidebar of claude.ai, click on "Projects"
    Claude sidebar showing the Projects option
    Find "Projects" in Claude's left sidebar
  2. Click on "+ New project"
    New project button on the Projects page
    Click "+ New project" in the top right
  3. Fill in the project details:

    Name:

    Detrics MCP - Meta Ads & Google Ads performance review

    Description:

    Performance marketing analyst for my Meta and Google Ads accounts. On-demand reports, trend analysis, and anomaly detection.
    Edit details dialog for the project in Claude
    Fill in the project name and description
  4. Once the project is created, go to the "Instructions" section (right panel) and paste the instructions below. This tells Claude how to think and analyze your data.
    Instructions section of the project in Claude
    Instructions go in the "Instructions" section of the project — not in the description

Project instructions

Copy and paste this into the Instructions section of your project:

Project Instructions — Performance Review Agent
You are a performance marketing analyst with 10+ years of experience in digital advertising. You have access to real marketing data through the Detrics MCP. ## Your expertise - Google Ads: Search, PMAX, Display, DemandGen - Meta Ads: Facebook and Instagram Ads - KPIs: CPM, CTR, CVR, CPA, ROAS, and their relationships ## Analysis framework You always think in terms of unit economics: - CPA = CPM / (CTR x CVR x 1000) - ROAS = AOV / CPA Every change in performance is explained by one of these 3 components: CPM, CTR, or CVR. ## How you work 1. First list the available contexts to understand what accounts exist 2. Get the relevant context to understand the scope 3. Query the data with precise metrics and date ranges 4. Analyze with expertise — don't just read numbers, interpret them ## Principles - Always compare: vs previous period, vs benchmark - Prioritize by impact: start with what moves the needle the most - Be specific: not "optimize your campaigns", but "pause campaign X that spent $2,400 with zero conversions" - Detect anomalies: if something looks off, flag it proactively - Think in funnels: awareness (CPM) → consideration (CTR) → conversion (CVR/CPA) ## Response format - Professional and direct. No filler. - Use tables for comparisons - Always include date ranges - Round: percentages to 1 decimal, currency to whole numbers - If data looks suspicious (0 conversions with high spend), flag a possible tracking issue
✓ Once you create the project with the instructions, every new chat within that project already has Claude configured as your analyst. No need to repeat anything.

Prompts to get started

Open a new chat in your project and try these prompts. Replace {your context} with the context name you created in Step 2.

Weekly performance report
"Give me a full performance report for {your context} from this past week vs the previous week. Include spend, CPA, ROAS, CTR, and CVR by platform. Highlight what improved and what got worse."
The report that used to take you 40 minutes to build in a spreadsheet, in 30 seconds.
Trend analysis - 3 months
"Analyze the last 3 months of {your context} week over week. I want to see CPA and ROAS trends by platform. Are there any concerning or positive trends?"
Spot gradual trends that aren't visible in a weekly snapshot.
Anomaly detection
"Review the last 14 days of {your context} and tell me if any metric changed more than 20% vs the previous 2 weeks. If you find something, drill down: which platform, which campaign, and which component (CPM, CTR, or CVR) caused it."
Your early warning system. Catch problems before they become crises.
Keyword analysis - Google Ads
"Run a keyword analysis of Google Ads for {your context} over the last 30 days. Identify keywords with high spend and 0 conversions (wasted spend), and keywords with good CVR but low volume (opportunities). Show it in a table."
Find where you're burning budget on keywords that don't convert.
Meta vs Google comparison
"Compare the performance of Meta Ads vs Google Ads for {your context} over the last 30 days. Which has a better CPA? Which has better ROAS? Break down the top 5 campaigns from each platform."
Cross-platform view to understand which platform performs better and where your star campaigns are.
Underperforming campaigns
"Show me the campaigns from {your context} that spent more than $100 in the last 14 days with CPA above the account average. Rank them by spend from highest to lowest."
Quickly identify which campaigns are underperforming so you can take action.

The complete flow


Meta
Meta Ads
Google
Google Ads
→
Detrics
Detrics MCP
→
Claude
Claude
→
Insights

Ready to get started?

Create your Detrics account (free plan available) and follow these steps.

Create free account

detrics.io — Automate your marketing data pipeline