Voice-of-customer aggregator (L3)

The steps

1. 01

   Centralize multi-channel feedback data

   Begin by identifying and establishing secure API or webhook connections to four primary feedback channels. For calls, use a conversation intelligence tool like Gong or Chorus. For tickets, connect Zendesk or Salesforce Service Cloud. For NPS, use Delighted or Vitally. For reviews, use G2 or TrustRadius via an aggregator like BrightLocal or a simple RSS feed. • Ownership: RevOps or Data Engineer. • Time Estimate: 4-6 hours. • Process: In your automation hub (Zapier, Make.com, or Workato), create a "New Feedback Entry" trigger for each source. Use the 'API Key' or 'OAuth' settings in each platform’s developer portal to authorize the connection. • Prompt/Config: Create a unified 'Raw Data' table in Airtable or Google BigQuery with fields for: [Source Name], [Raw Text/Transcript], [Customer Segment], [Date], and [Original URL]. • Pitfalls: Ensure you are pulling 'Full Transcripts' from call tools, not just summaries, as LLMs need the raw context to identify nuances. • Definition of Done: A centralized data repository actively populating with live streams from all four specified sources.

2. 02

   Clean and structure weekly data imports

   Set up an automated batch process to clean and prepare the raw text for LLM ingestion. This involves stripping PII (Personally Identifiable Information) and formatting different data types into a consistent structure. • Ownership: Data Analyst. • Time Estimate: 2-3 hours. • Process: Use a script or a tool like Clay to concatenate small snippets (like NPS) and chunk long transcripts (like 60-minute calls). • SQL/Formula: If using BigQuery: 'SELECT transcript_text, customer_id, timestamp FROM raw_feedback WHERE timestamp >= DATE_SUB(CURRENT_DATE(), INTERVAL 7 DAY)'. • Pitfalls: Don't ignore metadata! Ensure the customer's 'Plan Level' or 'ARR' is attached to each record so the LLM can weigh feedback from high-value accounts more heavily. • Definition of Done: A clean, formatted JSON or CSV export containing all feedback from the past seven days, ready for the LLM prompt.

3. 03

   Run thematic analysis with citations

   Configure an LLM (GPT-4o or Claude 3.5 Sonnet) to perform thematic analysis through a sophisticated system prompt. The prompt must mandate the identification of 3-5 recurring themes with direct citations. • Ownership: PMM or AI Specialist. • Time Estimate: 3 hours. • Example Prompt: 'Analyze the attached feedback from [Source A, B, C]. Identify the top 5 pain points or feature requests. For each theme, provide: 1) A 2-sentence summary, 2) The % of total feedback it represents, 3) 3 verbatim quotes with links to the source. Format as a markdown report.' • Pitfalls: Hallucinations. If you don't force 'Citations,' the LLM might invent complaints that sound plausible but don't exist in the data. • QA Check: Manually verify that the three 'Citations' in the first theme actually exist in the raw data export. • Definition of Done: Success occurs when the LLM outputs a structured report where every theme is backed by at least two distinct data sources.

4. 04

   Distribute digest to PMM and Product

   Automate the distribution of the 'Voice of Customer Digest' to the Product Management (PM) and Product Marketing (PMM) teams via Slack or Email every Monday at 9:00 AM. • Ownership: RevOps. • Time Estimate: 1 hour. • Process: In Zapier/Make, create a 'Scheduled Trigger' (Weekly). The action should be 'Send a Slack Block Kit Message' to the #product-insights channel. • Formatting: Use bold headers for themes. Include a direct link to the full Airtable/BigQuery record for anyone who wants to dive deeper. • Pitfalls: Sending this monthly. Monthly feedback is 'stale' for agile product teams. Stick to a weekly cadence to ensure feedback impacts the current or next sprint. • Definition of Done: A recurring Slack message or Email that successfully delivers the theme report to the leadership team without manual intervention.

5. 05

   Audit roadmap influence quarterly

   Establish a quarterly 'Accountability Sync' to bridge the gap between feedback and the product roadmap. This ensures that the VOC aggregator isn't just 'noise' but a driver of change. • Ownership: Head of Product & PMM. • Time Estimate: 1 hour per quarter. • Process: Audit the last 12 weekly reports. Cross-reference the 'Top 3 Themes' from the VoC aggregator against the 'Product Roadmap' in Jira or Productboard. • Analysis: Ask: 'Did we ship a fix or feature for any of the top themes identified by the AI?' If the answer is 'No' for two consecutive quarters, either the AI is surfaced the wrong insights, or the product team is ignoring the customer voice. • Pitfalls: Letting the meeting become a 'blame game.' Focus on whether the AI themes were 'Actionable' or 'Too Vague.' • Definition of Done: A documented 'Gap Analysis' showing which AI-detected themes were moved to the roadmap and which were deprioritized.

Next playbooks

• AI customer-interview synthesis (L3)
• AI SEO clusters with Clearscope/Frase (L3)
• AI-generated event follow-up (L3)
• Lifecycle email AI personalization (L3)