Skip to main content
BuildBetter automatically analyzes your calls, conversations, and imported data to extract key insights called Signals. These insights help you understand customer needs, track issues, and identify patterns without manual tagging.

​
What are Signals?

Signals are automatically extracted insights from your data including:
  • Customer feedback and feature requests
  • Bug reports and technical issues
  • Product complaints and pain points
  • Success stories and wins
  • Competitive mentions
  • Action items and decisions
  • Questions and objections

​
Signal Sources

Signals are automatically extracted from:

Call Recordings

Live conversations, meetings, and sales calls with full transcript analysis

Imported Conversations

Support tickets (Zendesk, Intercom, Kustomer), chat logs (Slack), and CRM data

Documents & Text

Uploaded files, meeting notes, and written feedback

External Integrations

CRM systems (Salesforce, HubSpot), surveys (Pendo, Typeform), and more

​
Automatic Extraction Process

1

Data Ingestion

Content is processed from calls, imports, and integrations
2

AI Analysis

Advanced language models analyze content for meaningful insights
3

Signal Classification

Insights are categorized by type, sentiment, severity, and other properties
4

Enrichment

Signals are enhanced with metadata, context, and entity associations
Extraction happens automatically when calls are processed or data is imported. No manual tagging required.

​
Signal Types

BuildBetter detects 35+ different signal types across three categories:

​
Universal Signals (Internal & External)

Detected in all conversations:
  • Improvement - Suggestions for enhancements
  • Complaint - Expressions of dissatisfaction
  • Issue - Problems or challenges mentioned
  • Inquiry - Questions or information requests
  • Compliment - Positive feedback or praise
  • Observation - Noteworthy comments or insights
  • Testimonial - Customer success stories
  • Idea - Creative suggestions or concepts
  • Feedback - General feedback or opinions
  • Competition - Mentions of competitors
  • Action Item - Tasks and next steps

​
Internal-Only Signals

Detected in internal conversations (team meetings, planning sessions):
  • Suggestion - Team improvement ideas
  • Decision - Key decisions made
  • Feature - Feature discussions
  • Strategy - Strategic planning points
  • Change - Process or product changes
  • Confusion - Areas of uncertainty
  • Concern - Worries or risks
  • Challenge - Obstacles identified
  • Opportunity - Growth opportunities
  • Achievement - Milestones reached
  • Milestone - Project progress markers
  • Update - Status updates
  • Priority - Priority discussions
  • Risk - Risk identification
  • Blockers - Impediments to progress
  • Customer Insight - Insights about customers
  • Dependency - Dependencies identified

​
External-Only Signals (Customer-Facing)

Detected in customer conversations:
  • Feature Request - Explicit feature asks
  • Bug - Software defects or errors
  • Objection - Sales or product objections
  • Discovery - Discovery phase insights
  • Question - Customer questions
  • Strategic - Strategic discussions
  • Interest - Product interest indicators
Signal types can be configured in Settings > Features > AI Labeling to enable/disable specific types for your organization.

​
Signal Properties

Each extracted signal includes:

​
Core Properties

  • Type: Classification from 35+ signal types
  • Content: The actual text/quote from the source
  • Summary: AI-generated concise summary
  • Source: Original recording, conversation, or import

​
Enrichment Properties

  • Sentiment: Score from -10 (very negative) to +10 (very positive)
  • Severity: Impact score from -10 to +10
  • Bias: External signal bias measurement
  • Emotions: Detected emotions (happiness, frustration, confusion, etc.)
  • Business Impact: Revenue, adoption, satisfaction, retention, efficiency implications

​
Context Properties

  • People: Associated contacts
  • Companies: Related organizations
  • Topics: Detected themes and subjects
  • Timestamp: Exact moment in source content
  • Personas: Customer persona associations
  • Tags: Organizational labels

​
Dynamic CRM Properties

Signals inherit metadata from connected systems:
  • Salesforce fields: Account info, opportunity data, custom fields
  • HubSpot properties: Contact properties, company data, deal info
  • Custom metadata: Organization-specific attributes
Use the properties panel when viewing a signal to see all extracted and enriched data.

​
Extraction Configuration

​
AI Labeling Settings

Access Settings > Features > AI Labeling to:
  • Enable/disable specific signal types
  • Configure extraction sensitivity
  • Set confidence thresholds
  • Customize type definitions for your domain

​
Extraction Methods

  • Automatic (default): AI extracts signals during processing
  • Manual: Users can create signals directly from transcript selections

​
Signal Quality

​
Quality Indicators

  • Confidence score for each signal
  • Source reliability rating
  • Extraction method (automatic vs manual)
  • Verification status (if using citation verification feature)

​
Improving Accuracy

  • Ensure complete, accurate transcripts
  • Use proper speaker labeling
  • Configure signal types for your industry
  • Provide feedback on incorrect extractions
Signal accuracy depends on transcript quality, speaker identification, and proper source configuration. Review important signals before acting on them.

​
Integration with Other Features

​
Signals Enable:

  • Document Generation: Create PRDs, reports from signal collections
  • Datasets: Organize signals for analysis with custom AI columns
  • Dashboards: Visualize signal trends and patterns
  • Workflows: Trigger actions based on signal detection
  • CRM Push: Send signals to Jira, Linear, Salesforce, HubSpot

​
Troubleshooting

  • Verify transcript is complete
  • Check speaker labeling accuracy
  • Review signal type configuration
  • Ensure extraction hasn’t been disabled for that type
  • Review confidence scores
  • Check context and full transcript
  • Verify signal type settings
  • Use manual reclassification if needed
  • Signals may appear in multiple related contexts
  • Filter by unique source or timestamp
  • Use deduplication in dataset views
Signal extraction runs automatically on all processed content. No manual action required to generate signals from your calls and data.