# Customer Feedback Routing Workflow

Automate feedback intake, deduplication, classification, evidence preservation, ownership, response, and product-learning loops.

## Prompt

You are a voice-of-customer automation architect who specializes in routing feedback without losing context or overcounting duplicates.

Inputs:
1. Feedback channels, formats, and customer segments: {{feedback_sources}}
2. Product taxonomy, teams, and decision processes: {{product_context}}
3. Urgency, severity, sentiment, and routing rules: {{classification_rules}}
4. CRM, support, product, analytics, and communication systems: {{systems}}
5. Privacy, response, service-level, and reporting constraints: {{constraints}}

Do the following:
1. Define a canonical feedback record that preserves source, customer context, verbatim text, permissions, product area, request type, impact, evidence, and relationship to an underlying issue.
2. Design ingestion, normalization, language handling, redaction, deduplication, clustering, confidence thresholds, and manual review for ambiguous classifications.
3. Route safety, accessibility, service failure, defect, feature request, usability, praise, and research interest to appropriate owners with deadlines and escalation.
4. Specify acknowledgement, customer follow-up, status changes, linked product decisions, closure, and rules preventing an automated promise of roadmap delivery.
5. Produce the data model, routing table, workflow states, response templates, deduplication logic, dashboard metrics, and test cases. Count affected customers and evidence separately from raw message volume.

## Best for

Product, support, and customer teams building a reliable feedback pipeline from multiple channels into accountable decisions.

## Compatible tools

- Claude
- ChatGPT

## How to use

- List every source and its available metadata.
- Define the product taxonomy with owners.
- Use explicit high-risk escalation rules.
- Preserve source context and communication ownership.

## Customization tips

- Separate duplicates from independent customer evidence.
- Set confidence thresholds for automated classification.
- Track decisions linked to themes.
- Never convert feedback into an automatic product promise.

## Example input

Sources: Support tickets, NPS comments, sales call notes, app feedback, and research interviews. Product: Fleet maintenance platform with web and mobile apps. Teams: mobile, scheduling, reporting, support, and safety. Rules: safety issues escalate within 15 minutes; confirmed defects create engineering issues; feature requests enter monthly review. Constraints: redact driver names, preserve source links, reply through the originating team, and never promise delivery dates automatically.

## Example output

The design creates one feedback record linked to multiple messages and affected accounts, preventing a repeated NPS comment and support thread from being counted as separate demand. Safety language routes to a human queue within 15 minutes regardless of sentiment score. Low-confidence product classification requires review. Customers receive source-channel acknowledgements, while roadmap status remains internal unless an approved communicator sends an update.