Intent Analysis

AI-powered intent analysis provides signals about what users are trying to accomplish, enabling intelligent campaign targeting.

Overview

Intent analysis classifies user behavior into actionable categories that help bidders match campaigns to user needs. The intent object provides the classification, confidence level, identified topics, and behavioral signals.

Intent Object Structure

{
  "intent": {
    "value": "purchase_inquiry",
    "confidence": 0.85,
    "topics": ["family travel", "outdoor activities", "golf"],
    "signals": ["asked_about_options", "price_sensitive"]
  }
}

Field Reference

Field	Type	Required	Description	Example
`value`	string	Yes	Intent classification	"purchase_inquiry"
`confidence`	number	Yes	Confidence score 0.0-1.0	0.85
`topics`	string[]	Yes	Identified topics from conversation	["family travel", "golf"]
`signals`	string[]	Yes	Intent signals detected in user behavior	["asked_about_options"]

Intent Values

purchase_inquiry

User is asking about buying, booking, or purchasing something.

Indicators:

Questions about pricing
Asking about availability
Requesting purchase information
Comparing options to buy

Example:

{
  "value": "purchase_inquiry",
  "confidence": 0.92,
  "topics": ["golf resort", "family vacation"],
  "signals": ["asked_about_pricing", "requested_availability"]
}

Use Cases:

Show product ads
Display pricing information
Highlight special offers
Include clear CTAs

information_seeking

User wants to learn more about a topic or get information.

Indicators:

"Tell me about..."
"What is..."
"How does..."
General questions

Example:

{
  "value": "information_seeking",
  "confidence": 0.88,
  "topics": ["golf equipment", "beginner tips"],
  "signals": ["asked_for_explanation", "seeking_recommendations"]
}

Use Cases:

Educational content
How-to guides
Product comparisons
Informational ads

support_request

User needs help with a problem or issue.

Indicators:

"I need help with..."
"How do I fix..."
Problem descriptions
Troubleshooting questions

Example:

{
  "value": "support_request",
  "confidence": 0.9,
  "topics": ["booking issue", "cancellation"],
  "signals": ["expressed_frustration", "needs_assistance"]
}

Use Cases:

Support services
Solution-oriented ads
Customer service offers
Problem-solving products

comparison

User is comparing different options or alternatives.

Indicators:

"Which is better..."
"Compare X and Y"
"What's the difference..."
Evaluation questions

Example:

{
  "value": "comparison",
  "confidence": 0.87,
  "topics": ["golf resorts", "vacation packages"],
  "signals": ["comparing_options", "evaluating_features"]
}

Use Cases:

Comparison charts
Feature highlights
Competitive advantages
Side-by-side comparisons

objection_handling

User has concerns, questions, or objections that need addressing.

Indicators:

"But what about..."
"I'm concerned that..."
Skeptical questions
Hesitation signals

Example:

{
  "value": "objection_handling",
  "confidence": 0.85,
  "topics": ["price", "quality", "reliability"],
  "signals": ["expressed_concern", "price_objection"]
}

Use Cases:

Testimonials
Guarantees
Risk reversal offers
Trust signals

Topics Array

The topics array contains identified topics from the conversation, extracted using NLP analysis.

Characteristics:

Lowercase, underscore-separated
Ordered by relevance
Typically 1-5 topics
Contextual to current conversation

Example:

{
  "topics": ["family travel", "outdoor activities", "golf", "beach resort"]
}

Use Cases:

Topic-based targeting
Keyword matching
Content relevance
Contextual advertising

Signals Array

The signals array contains behavioral signals detected in user interactions.

Common Signals:

Signal	Description
`asked_about_options`	User requested multiple options
`asked_about_pricing`	User inquired about prices
`price_sensitive`	User shows price consciousness
`quality_focused`	User emphasizes quality
`time_sensitive`	User has urgency
`comparison_shopping`	User comparing alternatives
`expressed_excitement`	Positive emotional signals
`expressed_concern`	Hesitation or worry
`asked_for_recommendations`	Seeking suggestions
`ready_to_purchase`	Strong buying signals

Example:

{
  "signals": ["asked_about_pricing", "price_sensitive", "comparison_shopping"]
}

Confidence Score

The confidence score indicates how certain the AI is about the intent classification.

Range: 0.0 to 1.0

Interpretation:

0.9 - 1.0: Very high confidence
0.8 - 0.9: High confidence
0.7 - 0.8: Moderate confidence
0.6 - 0.7: Low confidence
< 0.6: Very low confidence

Example:

{
  "confidence": 0.92
}

Best Practices:

Use higher confidence scores for critical decisions
Consider multiple signals for low confidence scores
Combine with other context signals

Complete Examples

Example 1: Purchase Intent

User asking about booking a golf resort:

{
  "ext": {
    "aura": {
      "messageSequenceNumber": 4,
      "messageLength": 52,
      "intent": {
        "value": "purchase_inquiry",
        "confidence": 0.92,
        "topics": ["golf resort", "family vacation", "booking"],
        "signals": [
          "asked_about_pricing",
          "requested_availability",
          "ready_to_purchase"
        ]
      },
      "sentiment": {
        "value": "positive",
        "score": 0.85
      }
    }
  }
}

Example 2: Information Seeking

User learning about golf equipment:

{
  "ext": {
    "aura": {
      "messageSequenceNumber": 2,
      "messageLength": 38,
      "intent": {
        "value": "information_seeking",
        "confidence": 0.88,
        "topics": ["golf clubs", "beginner equipment"],
        "signals": ["asked_for_explanation", "seeking_recommendations"]
      },
      "sentiment": {
        "value": "neutral",
        "score": 0.5
      }
    }
  }
}

Example 3: Comparison Shopping

User comparing vacation packages:

{
  "ext": {
    "aura": {
      "messageSequenceNumber": 5,
      "messageLength": 67,
      "intent": {
        "value": "comparison",
        "confidence": 0.87,
        "topics": ["vacation packages", "all-inclusive", "pricing"],
        "signals": [
          "comparing_options",
          "evaluating_features",
          "price_sensitive"
        ]
      },
      "sentiment": {
        "value": "neutral",
        "score": 0.6
      }
    }
  }
}

Publisher Implementation

Intent Detection

class IntentAnalyzer {
  async analyzeIntent(conversation) {
    const lastMessage = conversation[conversation.length - 1]

    // Use AI/ML model to classify intent
    const classification = await this.classifyIntent(lastMessage)

    // Extract topics using NLP
    const topics = await this.extractTopics(conversation)

    // Detect behavioral signals
    const signals = this.detectSignals(conversation)

    return {
      value: classification.intent,
      confidence: classification.confidence,
      topics: topics,
      signals: signals,
    }
  }

  detectSignals(conversation) {
    const signals = []
    const text = conversation
      .map(m => m.content)
      .join(' ')
      .toLowerCase()

    if (text.includes('how much') || text.includes('price')) {
      signals.push('asked_about_pricing')
    }

    if (text.includes('compare') || text.includes('vs')) {
      signals.push('comparison_shopping')
    }

    if (text.includes('recommend') || text.includes('suggest')) {
      signals.push('asked_for_recommendations')
    }

    return signals
  }
}

Usage Example

const analyzer = new IntentAnalyzer()

const conversation = [
  { role: 'user', content: 'I need a golf resort for my family' },
  { role: 'assistant', content: 'I can help with that...' },
  { role: 'user', content: 'How much does it cost?' },
]

const intent = await analyzer.analyzeIntent(conversation)
// {
//   value: 'purchase_inquiry',
//   confidence: 0.92,
//   topics: ['golf resort', 'family vacation'],
//   signals: ['asked_about_pricing']
// }

Bidder Implementation

Intent-Based Targeting

function shouldBid(bidRequest) {
  const intent = bidRequest.ext?.aura?.intent

  if (!intent) return false

  // Only bid on high-confidence purchase intent
  if (intent.value === 'purchase_inquiry' && intent.confidence > 0.8) {
    return true
  }

  // Bid on information seeking if topics match
  if (
    intent.value === 'information_seeking' &&
    intent.topics.some(t => ourTopics.includes(t))
  ) {
    return true
  }

  return false
}

Bid Adjustment

function calculateBid(baseValue, intent) {
  let modifier = 1.0

  // Increase bid for purchase intent
  if (intent.value === 'purchase_inquiry') {
    modifier = 1.5
  }

  // Increase for high confidence
  if (intent.confidence > 0.9) {
    modifier *= 1.2
  }

  // Increase for strong buying signals
  if (intent.signals.includes('ready_to_purchase')) {
    modifier *= 1.3
  }

  return basePrice * modifier
}

Topic Matching

function matchTopics(campaignTopics, intentTopics) {
  const matches = campaignTopics.filter(ct =>
    intentTopics.some(it => it.includes(ct) || ct.includes(it))
  )

  return matches.length / campaignTopics.length
}

Best Practices

1. Use Confidence Scores

Always check confidence before making critical decisions:

if (intent.confidence > 0.8) {
  // High confidence - proceed with intent-based targeting
}

2. Combine Multiple Signals

Don't rely on intent alone - combine with other signals:

function shouldShowAd(bidRequest) {
  const intent = bidRequest.ext.aura.intent
  const sentiment = bidRequest.ext.aura.sentiment

  return (
    intent.value === 'purchase_inquiry' &&
    intent.confidence > 0.8 &&
    sentiment.value === 'positive'
  )
}

3. Match Topics to Campaigns

Use topic matching for relevance:

const relevanceScore = matchTopics(campaign.topics, intent.topics)
if (relevanceScore > 0.5) {
  // Topics are relevant
}

4. Respect Low Confidence

Handle low confidence scores appropriately:

if (intent.confidence < 0.7) {
  // Use broader targeting or skip
  return false
}

Next Steps

Sentiment Analysis - Sentiment detection
Conversation Metadata - Operational metrics
Field Reference - Complete field documentation

Overview​

Intent Object Structure​

Field Reference​

Intent Values​

purchase_inquiry​

information_seeking​

support_request​

comparison​

objection_handling​

Topics Array​

Signals Array​

Confidence Score​

Complete Examples​

Example 1: Purchase Intent​

Example 2: Information Seeking​

Example 3: Comparison Shopping​

Publisher Implementation​

Intent Detection​

Usage Example​

Bidder Implementation​

Intent-Based Targeting​

Bid Adjustment​

Topic Matching​

Best Practices​

1. Use Confidence Scores​

2. Combine Multiple Signals​

3. Match Topics to Campaigns​

4. Respect Low Confidence​

Next Steps​

Overview

Intent Object Structure

Field Reference

Intent Values

purchase_inquiry

information_seeking

support_request

comparison

objection_handling

Topics Array

Signals Array

Confidence Score

Complete Examples

Example 1: Purchase Intent

Example 2: Information Seeking

Example 3: Comparison Shopping

Publisher Implementation

Intent Detection

Usage Example

Bidder Implementation

Intent-Based Targeting

Bid Adjustment

Topic Matching

Best Practices

1. Use Confidence Scores

2. Combine Multiple Signals

3. Match Topics to Campaigns

4. Respect Low Confidence

Next Steps