diff --git a/workflows/Analyze Google Forms feedback with Gemini AI and send weekly reports-13772/readme-13772.md b/workflows/Analyze Google Forms feedback with Gemini AI and send weekly reports-13772/readme-13772.md new file mode 100644 index 000000000..f9672eca9 --- /dev/null +++ b/workflows/Analyze Google Forms feedback with Gemini AI and send weekly reports-13772/readme-13772.md @@ -0,0 +1,108 @@ +Analyze Google Forms feedback with Gemini AI and send weekly reports + +https://n8nworkflows.xyz/workflows/analyze-google-forms-feedback-with-gemini-ai-and-send-weekly-reports-13772 + + +# Analyze Google Forms feedback with Gemini AI and send weekly reports + +This document provides a technical breakdown of the **Customer Feedback Analyzer** workflow, designed to automate the sentiment analysis and reporting of Google Forms responses using Gemini AI. + +--- + +### 1. Workflow Overview + +The workflow automates the transition from raw customer feedback to structured business intelligence. It monitors a Google Sheet (linked to a Google Form), applies generative AI to categorize and score the text, and distributes the findings across project management tools and email. + +**Logical Blocks:** +* **1.1 Data Collection:** Triggered on a schedule to fetch and filter new, unanalyzed responses. +* **1.2 AI Intelligence:** Uses Google Gemini to perform multi-dimensional analysis (sentiment, theme, and recommendations). +* **1.3 Data Persistence & Reporting:** Updates the source spreadsheet, logs high-level metrics in Notion, and sends a formatted HTML report via Gmail. + +--- + +### 2. Block-by-Block Analysis + +#### 2.1 Data Collection +**Overview:** This block acts as the entry point, ensuring only new data is processed to manage API costs and prevent duplicate analysis. +* **Nodes Involved:** `Daily feedback check`, `Read feedback responses`, `Filter unprocessed responses`. +* **Node Details:** + * **Daily feedback check (Schedule Trigger):** Configured to trigger every day at 9:00 AM. + * **Read feedback responses (Google Sheets):** Connects to the specific Document ID and Sheet Name where Google Form responses are stored. + * **Filter unprocessed responses (Code):** + * **Logic:** Iterates through rows. If the "Processed" column is empty or "no", it adds the row to an array. + * **Output:** It aggregates all individual responses into a single string (`feedbackText`) to allow the AI to process the entire batch in one call. + * **Edge Case:** If no new responses are found, it returns a `skip: true` flag to prevent downstream errors. + +#### 2.2 AI Intelligence +**Overview:** Transforms unstructured text into structured JSON data. +* **Nodes Involved:** `Analyze feedback with Gemini`, `Gemini Chat Model`. +* **Node Details:** + * **Analyze feedback with Gemini (Basic LLM Chain):** + * **Configuration:** Uses a system prompt requiring the AI to return a specific JSON schema containing `sentiment`, `score` (1-10), `category`, and `theme`. + * **Input:** Uses the `{{ $json.feedbackText }}` variable from the previous block. + * **Gemini Chat Model (Chat Google Gemini):** + * **Model:** `gemini-2.0-flash`. + * **Settings:** Temperature is set to `0.3` for consistent, deterministic outputs. + * **Failure Type:** API Authentication errors or context window limits if the feedback batch is excessively large. + +#### 2.3 Reporting & Distribution +**Overview:** Parses the AI's response and updates external systems. +* **Nodes Involved:** `Parse results and build report`, `Update feedback sheet with results`, `Log insights to Notion`, `Email report to team`. +* **Node Details:** + * **Parse results and build report (Code):** + * **Technical Role:** Uses Regex (`/\{[\s\S]*\}/`) to extract the JSON object from the AI's text response. + * **Logic:** Calculates the `positiveRate` and `negativeRate` and constructs a multi-section HTML string for the email body. + * **Update feedback sheet with results (Google Sheets):** + * **Operation:** `Update`. Maps the AI results back to the original rows using the `rowIndex`. + * **Log insights to Notion (Notion):** + * **Action:** Creates a database page. + * **Properties:** Maps `Total Responses`, `Positive Rate`, `Top Themes`, and `Recommendation`. + * **Email report to team (Gmail):** + * **Configuration:** Sends the generated `emailBody` as an HTML message. + * **Subject:** Dynamic subject line including the current ISO date. + +--- + +### 3. Summary Table + +| Node Name | Node Type | Functional Role | Input Node(s) | Output Node(s) | Sticky Note | +| :--- | :--- | :--- | :--- | :--- | :--- | +| Daily feedback check | Schedule Trigger | Workflow Entry | (None) | Read feedback responses | Data collection | +| Read feedback responses | Google Sheets | Data Retrieval | Daily feedback check | Filter unprocessed responses | Data collection | +| Filter unprocessed responses | Code | Data Filtering | Read feedback responses | Analyze feedback with Gemini | Data collection | +| Analyze feedback with Gemini | Basic LLM Chain | AI Processing | Filter unprocessed responses | Parse results and build report | AI analysis | +| Gemini Chat Model | Google Gemini | Language Model | (None) | Analyze feedback with Gemini | AI analysis | +| Parse results and build report | Code | Data Formatting | Analyze feedback with Gemini | Update feedback sheet, Log to Notion | Reporting | +| Update feedback sheet with results | Google Sheets | Data Persistence | Parse results and build report | Email report to team | Reporting | +| Log insights to Notion | Notion | Logging | Parse results and build report | (None) | Reporting | +| Email report to team | Gmail | Communication | Update feedback sheet | (None) | Reporting | + +--- + +### 4. Reproducing the Workflow from Scratch + +1. **Preparation:** + * Create a Google Sheet with headers: `Timestamp`, `Feedback`, `Processed`, `Sentiment`, `Category`, `Score`, `Theme`. + * Create a Notion Database with properties: `Date` (Title), `Total Responses` (Number), `Positive Rate` (Number), `Top Themes` (Text), `Recommendation` (Text). +2. **Trigger Setup:** Add a **Schedule Trigger** node set to "Daily". +3. **Data Fetching:** Add a **Google Sheets** node. Set the action to "Get Many" and point it to your feedback sheet. +4. **Batch Processing Logic:** Add a **Code** node. Paste logic to check if `row.Processed` is falsy. Create an object that combines the row index and the text response. +5. **AI Integration:** + * Add a **Basic LLM Chain** node. + * Connect a **Google Gemini Chat Model** node to its AI Model input. Select `gemini-2.0-flash`. + * In the Chain node, write a prompt asking for JSON output including: sentiment, score, and a summary recommendation. +6. **Data Parsing:** Add a **Code** node. Use `JSON.parse` on the AI output. Construct an HTML string using standard tags (`

`, `