n8n Workflow: Segment Donors & Personalize Emails (Sheets + Ollama)

Automatically segment your donors and personalize emails using Google Sheets + Ollama (Llama 3.1)

This n8n workflow reads donor data from your Google Sheets, validates and segments donors with JavaScript-based RFM + behavioral rules, then generates tailored donor messages using a local Ollama model (Llama 3.1)—and writes the results back to your “Donor Data” sheet.

What this workflow does

• Manually triggered donor processing: Run the workflow on demand to process all donor rows in your Google Sheets document.
• Normalize, deduplicate, and validate: It standardizes key fields (email, donation amount, donation date), deduplicates exact duplicate rows, and flags each row as valid or invalid.
• Track invalid rows: Invalid donor rows are appended or updated in a separate “Invalid Rows” sheet for follow-up.
• RFM + behavioral segmentation: For valid donors, it calculates RFM (recency, frequency, monetary) scores, assigns an RFM segment, and adds behavioral tags based on events attended and prior communications.
• Personalized prompt + local AI drafting: Each combined segment maps to a treatment plan (tone, ask type, channel, cadence), builds a per-donor prompt, and sends it to Ollama (Llama 3.1) to draft a short message.
• Quality checks + fallback: It validates the draft and falls back to a segment-based template when the model output fails checks.
• Writes back to Google Sheets: Updates each “Donor Data” row with the Personalized Message, segment tags, and review flags for tracking.

Use cases

• Produce donor outreach for fundraising campaigns from a live Google Sheets list without manual segmentation.
• Generate consistent, on-brand email content with local LLM support via Ollama.
• Maintain data quality by automatically routing problematic records to an “Invalid Rows” sheet.

Technical details (n8n + integrations)

• Manual Trigger to start processing.
• Google Sheets for reading and writing “Donor Data” and “Invalid Rows”.
• Core logic with if, code, and merge nodes (plus documentation via sticky note).
• Local AI generation using Ollama (Llama 3.1), with validation and template fallback.