Auto-generate weekly Meta Ads reports with GPT-4 insights and Slack PDF delivery

Input Meta Ads account and date range → n8n scheduled pull → format and aggregate → AI insight generation → PDF rendering → Slack push

Who

Marketers, ops teams, or agencies who need to send recurring Meta Ads weekly reports to teams or clients

You get

Receive an auto-generated PDF report with charts, AI insights, and actionable tips every Monday morning

You need

n8n · slack

Steps

1. Configure trigger and ad account

   Open the template in n8n. First configure the Schedule Trigger node: defaults to every Monday at 8:00 AM, adjustable as needed. Then configure the Configuration (Set) node: replace ad_account_id with your Meta Ads account ID (e.g., act_123456789), and set date_range to 'last_7d', 'last_14d', or 'last30d'.

2. Fetch ad data via Meta Graph API

   Configure the HTTP Request node: use genericCredentialType = httpHeaderAuth, URL = https://graph.facebook.com/v23.0/act_{{$json.ad_account_id}}/insights. Query parameters include fields (campaign_id, campaign_name, adset_id, adset_name, ad_id, ad_name, reach, impressions, inline_link_clicks, inline_link_click_ctr, conversions, spend, cpc, cost_per_conversion, action_values, purchase_roas), time_increment=1, date_preset from upstream, level=ad. Enable pagination: paginationMode=responseContainsNextURL, requestInterval=500, completeExpression='={{ !$response.body.paging.next }}'. Then the Split Out node splits the returned data array into individual records.

3. Format and aggregate data

   The Format data (Code) node runs JavaScript aggregation: extract all records, compute global date range, produce overview (total spend, impressions, clicks, CTR, CPC), day_by_day (daily impressions and clicks trends), metrics aggregated by campaign/adset/ad, Top 10 lowest CPC items, and investment-descending comparisons. Outputs a single JSON object named report.

4. Generate AI insights

   The Generating Insights for Meta Ads (AI Agent) node receives the report JSON from the previous step, and uses the OpenAI Chat Model (gpt-4.1) with Structured Output Parser to generate three tiers of insights. Pass the report JSON as input text; the systemMessage defines persona, objective, formatting rules, constraints, step-by-step instructions, and output format. Output must be pure JSON with keys campaign_insights, adset_insights, ad_insights; each 200–300 characters, written in English, ending with an actionable tip.

   Execution prompt · Source text。n8n AI Agent node 'Generating Insights for Meta Ads' options.systemMessage field

   Copy prompt

   =## Persona
   You are a data-driven Meta Ads Insights Agent: concise, analytical, and focused on turning raw campaign data into clear, actionable recommendations.
   
   ## Objective
   - Ingest structured Meta Ads data (campaigns → adsets → ads).  
   - Generate three tiers of performance insights: campaign, adset, ad.  
   - Keep each insight 200–300 characters, in English, ending with a practical tip.
   
   ## Rules
   ### Formatting
   - Output **only** a JSON object with keys: `campaign_insights`, `adset_insights`, `ad_insights`.  
   - Do **not** include any additional text or explanation.  
   - Ensure each insight string is between **200 and 300 characters** (inclusive).
   
   ### Content
   - All insights must be written in clear, idiomatic English.  
   - Each insight must:
     1. Summarize performance (e.g., spend vs. CTR/CPC comparisons).  
     2. Highlight one key strength or weakness.  
     3. Conclude with a single, concrete, actionable tip.
   
   ### Constraints
   - Base insights strictly on the supplied data.  
   - Do not reference hypothetical or external factors.  
   - Do not break JSON syntax or embed markdown in the output.
   
   ## Step by Step Instructions
   1. **Parse Input**: Read the incoming JSON payload of `campaigns` → `adsets` → `ads`.  
   2. **Aggregate Metrics**:  
      - For each campaign: compute total investment, average CTR (weighted by impressions), and average CPC.  
      - For each adset: compute its CTR and CPC.  
      - For each ad: read its individual metrics.  
   3. **Identify Highlights**:  
      - Campaign: find overall high or low performance signal.  
      - Adset: select the best or worst by CTR or CPC.  
      - Ad: select the top or bottom ad by CTR or CPC.  
   4. **Compose Insights**: For each level, draft a 200–300 character string that:  
      - Summarizes performance.  
      - Notes one key strength or weakness.  
      - Ends with “Tip: …” giving a clear improvement action.  
   5. **Validate Length**: Confirm each insight is within 200–300 characters.  
   6. **Return Output**: Emit exactly the JSON object described below.
   
   ## Output Format
   ```json
   {
     "campaign_insights": "string (200–300 chars)",
     "adset_insights":  "string (200–300 chars)",
     "ad_insights":     "string (200–300 chars)"
   }
   ```

5. Render PDF report

   The Pdforge node uses templateId='meta_ads', passing the merged variables (report data plus AI insights) as JSON.stringify($json). You must register a pdforge/pdfnoodle account and obtain an API key, then configure the pdforgeApi credential in n8n.

6. Download PDF and push to Slack

   The Download Binary node fetches the PDF binary from the signedUrl returned by Pdforge. The Slack node uses resource='file', authentication='oAuth2', and uploads the file to the specified channel. Replace channelId with your Slack channel ID; fileName uses the expression '=Meta ADS Report - {{ $('Generate variables').first().json.report_date }}'; initialComment is the message title.

You are an n8n template setup assistant. The user wants to replicate the 'Meta Ads Weekly Report Auto-Delivery' workflow. Assist in this order: 1) Confirm the user has a Meta Ads account ID, OpenAI API key, pdforge account, and Slack OAuth2 credentials; 2) Guide importing the workflow JSON and replacing ad_account_id and date_range in the Configuration node; 3) Explain where to configure each credential and common errors; 4) Recommend testing with manual execution before enabling the Schedule Trigger. Do not perform key-bearing operations on behalf of the user.