Claude Code for Marketing: Connect Your Data and Run a Repeatable Campaign Review

What can Claude Code do for marketing?

Claude Code combines an AI conversation with tools and local files. Through Model Context Protocol, or MCP, it can access connected services. Its project instructions can also preserve the conventions you want it to follow across reviews. Claude Code MCP documentation, project instructions documentation.

For marketing teams, that supports a practical workflow:

Claude Code can then organise the findings into a local report, keep your review instructions alongside it and compare the next review with your previous decisions.

For a wider set of use cases, see our guide to Claude for marketing.

Connect your marketing data to Claude Code

Start by connecting the accounts you want to review in SourcetoMCP. Check that you have selected the correct advertising account, website or property.

Open Connect Agent in your SourcetoMCP dashboard and select Claude Code. Copy the command shown there. It follows this structure:

claude mcp add --transport http sourcetomcp YOUR_MCP_SERVER_URL --scope user

Replace YOUR_MCP_SERVER_URL with the URL supplied by your dashboard.

Start Claude Code, run /mcp, select SourcetoMCP and choose Authenticate. Complete the browser sign-in and approval process.

Your first request should establish what is available:

List my connected marketing accounts and the reporting operations available for them. Identify each account by name and ID. Tell me if anything needs reconnecting before we start the review.

This helps prevent a familiar reporting mistake: analysing the wrong account because two connections have similar names.

Create a brief you can reuse

Create a folder for your campaign reviews. Add a CLAUDE.md file with the business context and reporting rules you want Claude to follow.

You can ask Claude Code to create it from a brief like this:

Purpose: Review acquisition campaign performance.

Business context:
- Product: [your product]
- Main conversion: [your defined conversion]
- Target CPA: [amount and currency]
- Markets: [countries]
- Accounts: [approved account names and IDs]

Review rules:
- Compare equal-length periods with matching weekdays.
- State the exact dates and account timezone where available.
- Keep currencies and platform attribution separate.
- Calculate ratios from totals.
- Report missing data and incomplete result sets.
- Distinguish observed changes from possible explanations.
- Recommend account changes for review before applying them.

Output:
- Main findings
- Supporting figures
- Questions still unresolved
- Three prioritised next actions

Make the conversion definition specific. A form submission, qualified lead and completed purchase represent different business outcomes.

The target CPA is your business benchmark. Claude should use the number you provide and explain when the available conversion data does not match that benchmark.

Run a consistent campaign comparison

Start with a narrow review of one account:

Review my selected Google Ads account. Compare the latest seven complete days with the preceding seven days. State the dates, currency and timezone where available. Report spend, impressions, clicks, conversions, CTR, CPC and CPA by campaign. Calculate ratios from totals. Flag conversion reporting delays, missing values and any truncated results. Identify the three changes that deserve investigation.

Hypothetical example:

The account spent 10% more and recorded 20% fewer conversions. Together, those changes increased CPA by 37.5%.

That gives you a useful starting point. The next question is which campaigns account for the deterioration.

Ask:

Identify the campaigns contributing most to the CPA increase. Show their spend and conversion changes separately. Investigate device performance and available search-term data. Explain which findings are supported by the reports and which need further checking.

SourcetoMCP’s Google Ads search-term operation excludes Performance Max data. Keep that limitation visible when interpreting a mixed campaign account.

Turn observations into decisions

A review becomes more useful when every recommendation includes evidence and a next step.

For example:

Ask Claude to separate three things: the measured change, its possible explanation and the action you will use to test that explanation.

Avoid combining conversions reported by different advertising platforms into a single customer total. Attribution can overlap, and the platforms may use different conversion definitions.

Save a review you can act on next week

Once you have checked the findings, ask Claude Code to save the report:

Save this review as a Markdown file in my campaign-reviews folder. Include the account IDs, reporting periods, metric definitions, calculations, limitations and three next actions. Add an owner and review date placeholder for each action.

At the next review, ask it to read the previous report before retrieving fresh data.

That lets you revisit whether an action happened, whether the relevant metric changed and whether the original explanation still fits the evidence.

Do you need coding experience?

This workflow uses a terminal command for setup and plain-language requests for the analysis. You should be comfortable checking account access, reviewing calculations and managing the resulting files.

The marketing judgement still comes from your business context: what a conversion is worth, which campaigns matter and how much evidence you need before changing spend.

Connect your first account through SourcetoMCP, then run one campaign review with a defined period, conversion goal and output format.