Astra Forge
Astra Forge
I'm going to walk you through the exact n8n workflow I built to automate client reporting for a small marketing agency. It runs every month, pulls data from all the major platforms, generates a plain-English summary, and emails the report — without a human doing anything except reviewing it.
If you want to skip straight to the template, the downloadable workflow is at /guide. But read this first so you understand what it's doing and why.
n8n is an open-source workflow automation tool. Think of it like Zapier, but you can self-host (we use Hostinger) it, there's no per-task pricing, and it's actually powerful enough to handle complex multi-step logic without duct tape.
For agencies, the value proposition is specific:
The combination of those four things is what makes it the right tool here. Other options either cost too much at scale, require code, or can't handle the multi-step logic of a real reporting workflow.
The full workflow has five stages. Here's what each one does:
A Schedule node fires on the first of each month at 8am. It passes the previous month's date range to all downstream nodes as variables (e.g., startDate: 2026-02-01, endDate: 2026-02-28).
Parallel branches fetch data simultaneously from each source. GA4 node pulls sessions, users, and top pages. Google Search Console node pulls clicks and impressions by query. Meta Ads node pulls spend, reach, and CPA. Each branch outputs a structured JSON object with the key metrics.
A Merge node combines all the data into a single object. A Code node (about 15 lines of JavaScript) calculates month-over-month percentage changes and formats the numbers into human-readable strings.
The formatted data goes into an AI node (Claude or GPT-4o). The system prompt instructs it to write a 3-paragraph executive summary in plain English, noting what improved, what declined, and what to focus on next month. Output is clean prose, no bullet points.
A Gmail node sends the completed report to the client. Optionally, a Google Sheets node logs the raw data for historical tracking. If you prefer Slack, swap Gmail for a Slack node — same data, different delivery.
If you're new to n8n, these are the nodes doing the heavy lifting in this workflow:
The email the client receives looks like a report a competent junior analyst wrote. It opens with the headline numbers, gives a paragraph on what drove the results, notes any anomalies, and ends with a recommendation for next month.
It's not generic. The AI narrative is grounded in the actual numbers, so it says things like "Organic traffic declined 12% month-over-month, primarily in the blog section — this aligns with the indexing issue we flagged mid-February" rather than "Traffic was mixed this month."
The quality of the narrative depends on the quality of the prompt. Spend 20 minutes writing a good system prompt that knows your client's goals and industry. You write it once; it applies to every report after that.
Running this for 6 clients doesn't mean building 6 workflows. The cleaner approach is one workflow with a client configuration table — a Google Sheet or Airtable with one row per client, storing their GA4 property ID, ad account IDs, email address, and any custom instructions for the AI prompt.
The workflow loops through the table, runs the full reporting logic for each row, and sends each client their own email. Total runtime for 6 clients: under 15 minutes. Total human time: reviewing 6 draft reports and clicking send.
The free guide at /guide includes the complete n8n workflow JSON file, ready to import. It includes the GA4 connection, the AI narrative node with a tested system prompt, and the Gmail delivery. You'll need to add your own API credentials, but the structure is there — no building from scratch.
The complete workflow JSON — import it, add your credentials, and your reporting is automated. No coding required.
Get the Free Guide →