How to Set Up MCP for Marketing Analytics in 15 Minutes

What you'll have at the end: An AI agent analyzing your Google Ads, GA4, and ad spend together — with specific findings delivered in 24 hours. It takes 15 minutes to set up, and the first growth insight usually lands before your next meeting.

This guide is for marketers who are already sold on MCP marketing and want the practical steps. No engineers. No BigQuery setup. No custom code. Just your accounts and 15 minutes.

Prerequisites

Before you start, make sure you have:

A GA4 property with at least 30 days of data
At least one ad platform account (Google Ads or Meta — pick whichever has more spend)
Optionally: Stripe or Shopify for revenue data (strongly recommended — this is what unlocks real ROI analysis)
A Cogny Solo account ($9/month — or start free with 15 calls to verify it works before committing)

That's it. No API keys to generate, no OAuth apps to configure, no whitelists to update. Cogny handles the authentication flow for each platform.

Step 1: Connect GA4

GA4 is the right starting point because it's the source of truth for what happens after a click. Every channel's performance ultimately flows through here.

Log into Cogny and open your workspace
Navigate to Integrations (left sidebar)
Click Add integration → Google Analytics 4
Sign in with the Google account that has GA4 access
Select your property from the dropdown

That's the whole step. Cogny connects via the GA4 Data API — the same API Google's own tools use. No sampling, no delays.

What you get immediately: The AI can now query sessions, conversions, traffic sources, landing page performance, and goal completions going back as far as your property history. When you ask "which channels drove the most conversions last 90 days," this data answers it.

Step 2: Connect Your Primary Ad Platform

Connect whichever platform has the most spend first — Google Ads or Meta. You can add the other one in the same session, but one is enough to see the first round of insights.

Google Ads

In Integrations, click Add integration → Google Ads
Authorize with the Google Ads account (MCC or direct account)
Select the account(s) to connect

Cogny pulls campaign performance, keyword data, ad group structure, Quality Scores, impression share, and auction insights. This is where budget waste shows up — keywords bidding against each other, campaigns with poor structure, ROAS by campaign type.

Meta Ads

In Integrations, click Add integration → Meta Ads
Log in with Facebook Business Manager access
Select the ad accounts to connect

Cogny pulls campaign performance, ad set targeting, creative performance, frequency data, and audience overlap. Frequency problems and creative fatigue are the most common Meta findings in the first 24 hours.

Pick one, connect it, move on. You can always add more later. The AI compounds value as you add more connections — but it starts delivering insights from two sources.

Step 3: Connect Your Revenue Data

This is the step most setups skip — and the one that unlocks real ROI analysis.

Without revenue data, your AI agent can tell you which channels drive conversions, but it can't tell you which channels drive revenue. A campaign with 500 conversions at $50 CPA and a campaign with 200 conversions at $80 CPA look different on cost alone — but if the second campaign's customers have 3× higher LTV, the $80 CPA is the better business.

Stripe

In Integrations, click Add integration → Stripe
Paste your Stripe restricted API key (read-only — Cogny only reads, never writes)
Select the date range to pull

Now the AI can correlate GA4 traffic with Stripe revenue events: which sources drive customers who actually pay, average order value by channel, refund rates by acquisition source.

Shopify

In Integrations, click Add integration → Shopify
Enter your store URL and follow the OAuth flow
Cogny requests read-only access to orders, products, and customers

Shopify gives the AI order-level revenue data: revenue by channel, product performance, customer acquisition cost by cohort.

Don't have either? Connect whatever revenue source you have — WooCommerce, BigQuery with your own data warehouse, or a custom webhook. Cogny supports all of them. If you have no revenue data yet, the GA4 + ads combination still delivers useful findings — you're just working with conversion counts rather than revenue figures.

Step 4: What to Expect in the First 24 Hours

Once you have at least two sources connected, start a chat with your AI agent. Ask something concrete:

"What's my blended ROAS across Google and Meta for the last 90 days? Where am I overpaying?"

The AI pulls live data from all connected sources and responds with specific findings, not generic observations.

Typical first-24-hour findings include:

Budget inefficiency by campaign type. "Your branded search campaigns are running 22% over the recommended budget cap while Performance Max is underfunded relative to its ROAS."
Cross-channel attribution gaps. "17% of conversions credited to direct traffic in GA4 correspond to Meta click windows — your last-click model is understating Meta's contribution."
Creative fatigue signals. "Three of your top five Meta ad creatives have frequency above 4.5 in the last 14 days, with CTR declining 40% from baseline. This is suppressing delivery quality."
Quick structural wins. "Two exact-match keyword groups in your Google Ads account are competing on the same queries. Consolidating them would eliminate impression share loss."

These aren't AI hallucinations — they're specific observations derived from your actual numbers. The agent cites which data it pulled and from which platform.

Step 5: Act on Your First Growth Ticket

When the AI surfaces a finding with a clear action, it generates a growth ticket — a structured recommendation with:

The specific problem identified
The proposed action
The expected outcome (with a measurable hypothesis)
How to track whether it worked

Your job at this point is straightforward: approve it, execute it, and mark it done.

Example ticket:

Finding: Branded campaign CPCs have risen 34% over 90 days while impression share has declined. Competitor analysis shows two new entrants bidding on your brand terms.

Action: Add negative keywords for competitor misspellings, adjust bid strategy from Target CPA to Target ROAS at 400%, set bid floor at $0.80 to prevent below-threshold impressions.

Expected outcome: CPCs return to prior range within 14 days; impression share stabilizes above 85%.

Track via: Google Ads weekly performance report, compare CPC and impression share to 90-day baseline.

You review this, make the changes in Google Ads (takes 5 minutes), and the AI tracks the outcome against the hypothesis over the following weeks. Over time, this builds an organizational memory of what works for your specific business — not just what works generically.

DIY MCP vs. Cogny

If you're technically minded, you might be considering building this yourself. Here's what that actually involves:

What you need	DIY	Cogny
Engineer time	40–80 hours to build the initial stack	Zero
BigQuery instance	Configure, maintain, pay for storage + queries	Included
Custom MCP agents	Write, test, deploy, host, monitor	Built in
Platform API credentials	Generate, rotate, whitelist, maintain	Handled
Ongoing maintenance	API version changes, authentication refreshes, model updates	Zero
Time to first insight	2–6 weeks (if nothing breaks)	Under 24 hours

The technical pieces are not complicated in isolation — it's the integration surface and ongoing maintenance that erodes the value. Every marketing platform updates its API on its own schedule. Every authentication token has an expiry. Every new AI model requires prompt adjustments.

Cogny handles all of that continuously so you don't have to. The 15-minute setup isn't marketing copy — it's the actual time to first insight because we've already built and maintained the infrastructure stack.

For a full breakdown of the technical comparison, see MCP marketing overview.

Frequently Asked Questions

Do I need a developer to set up MCP for marketing?

No. Cogny's integrations are OAuth flows — the same as connecting any SaaS tool. No API keys, no code, no infrastructure to configure. If you can connect HubSpot to Salesforce, you can set up Cogny.

How long does it take to see the first insights?

Once you have two sources connected, you can start a chat immediately. Most users get their first specific finding within the first session (minutes, not hours). The "24 hours" framing refers to having enough session history for pattern detection on your specific account.

Is my marketing data secure?

Cogny reads data via read-only API connections — the AI cannot make changes to your ad accounts. All connections use OAuth 2.0 with the minimum required scopes. Data is processed in-memory for analysis and not stored in a third-party data warehouse.

Can I connect more than one ad account?

Yes. Cogny supports multiple accounts per platform and multiple platforms simultaneously. You can connect your Google Ads MCC (and all accounts underneath it), Meta Business Manager with multiple ad accounts, and multiple GA4 properties in the same workspace.

What if I only have GA4 right now?

Start anyway. A single connected source gives you immediate access to GA4 analysis — traffic patterns, landing page performance, conversion paths, goal funnel drop-off. It's not the full picture, but it's more than most teams have queried coherently. Add ad platforms as you go.

Get Started

Connect your first source in 5 minutes. The free tier gives you 15 AI calls to verify it works — no credit card required until you see the value.

Cogny Solo — $9/month → — Full MCP endpoint access, all platform integrations, conversational analysis, unlimited chat
Cogny Cloud — $499/month → — Autonomous growth tickets, truth ledger, agent execution, 2,000 AI credits/month

The ROI math is simple: if your AI agent finds one optimization that saves $500/month in wasted ad spend, the tool pays for itself in 18 days. Most accounts find that within the first session.

Cheaper than a coffee subscription. Faster than hiring an analyst.