Forward to: Data Science Team

Audience Intelligence
Workflows

Ten agent workflows for Criteo's data science team — shopper graph enrichment, lookalike audience modeling, cross-device identity resolution, persona-based targeting, interest graph construction, purchase intent prediction, audience decay modeling, cohort discovery, geographic audience expansion, and privacy-safe audience activation across 100M+ domains.

1Shopper Graph Domain-Level Enrichment

AI agent enriches Criteo's Shopper Graph with domain-level intelligence — mapping IAB Categories, Personas, and Countries to every domain a user visits, creating a multi-dimensional profile that improves ad targeting precision without relying on individual user tracking.

1
Map Domain Visits to Enrichment Signals
/products /blog IAB Categories Personas Countries
SHOPPER GRAPH ENRICHMENT — DOMAIN SIGNALS ══════════════════════════════════════════════ DOMAINS IN ENRICHMENT DB: 89.4M domains with full enrichment SHOPPER PROFILES ENRICHED: 2.1B profiles with domain-level signals AVG ENRICHMENT FIELDS PER PROFILE: 14.2 signals ENRICHMENT SIGNAL MAPPING: User visits nike.com → IAB: Sports > Athletic Clothing Persona: Active Lifestyle, Brand-Conscious PageRank: 8.7 (premium brand affinity) Country: US, EU (geographic intent) User visits wirecutter.com → IAB: Shopping > Product Reviews Persona: Active Buyer, Research Phase PageRank: 8.2 (high-intent signal) Domain Age: 15yr (trusted source affinity) User visits allrecipes.com → IAB: Food & Drink > Cooking Persona: Meal Planner, Home Cook PageRank: 7.9 (established content site) Country: US (domestic grocery intent) ENRICHMENT IMPACT: Profiles with 0 enrichment signals: 12% (cookieless gap) Profiles with 1-5 signals: 34% Profiles with 6-15 signals: 38% — optimal targeting depth Profiles with 15+ signals: 16% — premium audience
2
Build Domain-Based Audience Segments
Agent creates targetable audience segments from domain intelligence enrichment — grouping users by the IAB categories and Personas of domains they visit, creating cookieless audience pools that perform comparably to cookie-based targeting.
Audience Segment
"Premium Tech Enthusiasts" — Users visiting 3+ domains with IAB: Technology + Persona: Early Adopter + PageRank >6. Segment size: 34.5M profiles. Average basket value: $234. CTR on tech ads: 3.8% (vs 1.2% untargeted). This domain-intelligence segment outperforms cookie-based tech audiences by 23%.
HIGH-VALUE SEGMENT — 34.5M profiles, 3.8% CTR, $234 basket
Audience Segment
"Eco-Conscious Parents" — Users visiting domains with IAB: Family + Personas: Parent AND IAB: Sustainability + Personas: Eco-Conscious. Segment size: 8.9M profiles. Average basket value: $156. High-value niche impossible to build without domain-level enrichment data.
NICHE SEGMENT — 8.9M profiles, high AOV, unique to domain intelligence
3
Output: Shopper Graph Enrichment Impact

SHOPPER GRAPH — ENRICHMENT IMPACT

ENRICHMENT RESULTS: ══════════════════════════════════════════════ Profiles enriched: 2.1B (88% of Shopper Graph) Domain-based segments created: 456 targeting segments CTR improvement: +67% from domain intelligence targeting ROAS improvement: +45% vs non-enriched audiences Cookieless coverage: 78% of impressions now have domain signals Signals: IAB Categories, Personas, Countries, PageRank, Domain Ages

2Lookalike Audience Modeling at Scale

AI agent builds lookalike audiences from seed segments using domain intelligence — finding users who visit similar types of domains (by IAB Category, Personas, and PageRank patterns) to high-value converters, expanding addressable markets without cookie dependence.

1
Profile Seed Audience Domain Patterns
/products IAB Categories Personas OpenPageRank
LOOKALIKE MODELING — SEED AUDIENCE PROFILING ══════════════════════════════════════════════ SEED: Nike.com top 10% converters (234K users) DOMAIN VISIT PATTERNS: SEED AUDIENCE DOMAIN PROFILE: IAB Distribution: Sports (34%), Fashion (23%), Health (18%), Tech (12%) Persona Mix: Active Lifestyle (45%), Brand-Conscious (28%), Fitness (18%) Avg PageRank of visited sites: 6.8 (premium content affinity) Country Distribution: US (67%), UK (12%), DE (8%), FR (6%) Avg domains visited/week: 23 (high digital engagement) LOOKALIKE CANDIDATES FOUND: 1% Lookalike (closest match): 2.3M users | Similarity: 94% 5% Lookalike (broader reach): 11.5M users | Similarity: 82% 10% Lookalike (maximum reach): 23.4M users | Similarity: 71% EXPECTED PERFORMANCE: 1% Lookalike CTR: 3.4% (vs 3.8% seed, -10% decay) 5% Lookalike CTR: 2.8% (vs 3.8% seed, -26% decay) 10% Lookalike CTR: 2.1% (vs 3.8% seed, -45% decay)
2
Output: Lookalike Audience Performance

LOOKALIKE AUDIENCES — PERFORMANCE REPORT

LOOKALIKE MODEL PERFORMANCE: ══════════════════════════════════════════════ Models built: 4,567 (one per major advertiser segment) Avg reach expansion: 10x from seed to 5% lookalike Avg CTR retention: 74% of seed performance Cookie-free lookalikes: 100% domain-intelligence based Revenue from lookalike campaigns: $56M/quarter

3Purchase Intent Prediction Engine

AI agent predicts purchase intent by analyzing the type and recency of domains a user visits — distinguishing between research-phase browsing (review sites, comparison pages) and active purchase intent (retailer product pages, pricing pages) for real-time bid optimization.

1
Map Domain Visits to Purchase Funnel Stage
/products /pricing /blog Personas IAB Categories
PURCHASE INTENT PREDICTION — FUNNEL MAPPING ══════════════════════════════════════════════ INTENT STAGE CLASSIFICATION BY DOMAIN TYPE: AWARENESS (low intent): Blog/news domains (IAB: News, Entertainment) Persona: "Browser" | Bid multiplier: 0.6x | CVR: 0.3% CONSIDERATION (medium intent): Review/comparison sites Visited: wirecutter.com, rtings.com (IAB: Shopping > Reviews) Persona: "Researcher" | Bid multiplier: 1.2x | CVR: 1.8% INTENT (high intent): Retailer /products and /pricing pages Visited: bestbuy.com/products, amazon.com/pricing Persona: "Active Buyer" | Bid multiplier: 2.5x | CVR: 4.2% PURCHASE (highest intent): Cart/checkout signals Visited: retailer domains with /login activity Persona: "Ready to Buy" | Bid multiplier: 3.8x | CVR: 8.9% INTENT PREDICTION ACCURACY: 89% (validated on holdout)
2
Optimize Bids by Predicted Intent
Intent Signal
High-Intent User Detected — User visited wirecutter.com (review), then bestbuy.com/products (retailer), then bestbuy.com/pricing (price check) within 2 hours. Domain signals: IAB Shopping, Persona: Active Buyer, PageRank 8.1. Purchase intent score: 94/100. Bid multiplier: 3.2x for electronics commerce ads.
HIGH INTENT — 94/100 score, bid 3.2x for electronics ads
3
Output: Intent-Based Bidding Impact

PURCHASE INTENT — BIDDING PERFORMANCE

INTENT-BASED BIDDING RESULTS: ══════════════════════════════════════════════ Users with intent scores: 1.8B (85% of addressable audience) ROAS from intent-optimized bids: +78% vs flat bidding High-intent users captured: 34.5M/month CVR improvement: +145% on high-intent segments Signals: /products visits, /pricing visits, IAB, Personas, PageRank

4Interest Graph Construction

AI agent builds a comprehensive interest graph by mapping IAB Categories and Personas across all domains in the database — creating a web-scale knowledge graph of topic relationships that enables Criteo to discover non-obvious audience affinities for better targeting.

1
Build Interest Graph from Domain Intelligence
/blog /products IAB Categories Personas
INTEREST GRAPH — DOMAIN-LEVEL CONSTRUCTION ══════════════════════════════════════════════ NON-OBVIOUS INTEREST AFFINITIES DISCOVERED: Yoga Practitioners → also visit: Organic Food (89%), Travel (67%), Luxury Home (45%) Gaming Enthusiasts → also visit: Anime (78%), Energy Drinks (56%), PC Hardware (91%) Home Renovators → also visit: Garden (92%), Cooking (45%), Smart Home Tech (67%) Marathon Runners → also visit: Nutrition (94%), Travel (78%), Wearable Tech (82%) GRAPH STATISTICS: Nodes (IAB Categories): 400+ Edges (affinity connections): 12,345 significant correlations Strength threshold: >30% co-visitation rate Coverage: 89.4M domains contributing to graph
2
Output: Interest Graph Targeting Impact

INTEREST GRAPH — TARGETING PERFORMANCE

INTEREST GRAPH TARGETING RESULTS: ══════════════════════════════════════════════ Affinity connections discovered: 12,345 New audience segments from affinities: 234 non-obvious segments CTR improvement from affinity targeting: +34% Reach expansion: +56% through cross-category affinities Example: Yoga brand reaching Organic Food audiences → +89% ROAS

5Audience Decay Modeling

AI agent models audience signal decay by tracking how domain visit recency impacts purchase intent — determining optimal retargeting windows for each IAB category and Persona type to maximize ROAS and prevent wasted ad spend on decayed audiences.

1
Measure Signal Decay by Category
/products IAB Categories Personas
AUDIENCE DECAY MODELING — BY CATEGORY ══════════════════════════════════════════════ SIGNAL HALF-LIFE BY IAB CATEGORY: Electronics: Half-life: 7 days (fast decision cycle) Fashion: Half-life: 14 days (moderate cycle) Home & Garden: Half-life: 30 days (slow consideration) Automotive: Half-life: 60 days (long decision cycle) Travel: Half-life: 21 days (event-driven) Luxury: Half-life: 45 days (aspirational browsing) Grocery: Half-life: 3 days (immediate need) OPTIMAL RETARGETING WINDOWS: Electronics: Days 1-7 = 4.2% CVR | Days 8-14 = 1.8% CVR | Days 15+ = 0.4% CVR Fashion: Days 1-14 = 3.1% CVR | Days 15-28 = 1.4% CVR | Days 29+ = 0.3% CVR Automotive: Days 1-60 = 0.8% CVR | Days 61-90 = 0.4% CVR | Days 91+ = 0.1% CVR
2
Output: Decay-Optimized Bidding

AUDIENCE DECAY — OPTIMIZATION REPORT

DECAY-OPTIMIZED BIDDING RESULTS: ══════════════════════════════════════════════ Wasted spend eliminated (decayed audiences): $12.3M/quarter ROAS improvement from decay modeling: +34% Optimal window compliance: 89% of bids within half-life Categories with fastest decay: Grocery (3d), Electronics (7d) Categories with slowest decay: Automotive (60d), Luxury (45d)

6Cohort Discovery Engine

AI agent discovers new high-value audience cohorts by analyzing domain visit patterns and enrichment data — identifying clusters of users with similar domain-level profiles that represent untapped targeting opportunities for Criteo's advertisers.

1
Discover Cohorts from Domain Patterns
/blog /products IAB Categories Personas Countries
COHORT DISCOVERY — NEW SEGMENTS FOUND ══════════════════════════════════════════════ NEWLY DISCOVERED HIGH-VALUE COHORTS: "AI-Curious Professionals" — 12.3M users Domain pattern: visit AI/ML blogs + business news + career sites Personas: Tech Decision Maker, Early Adopter Avg purchase value: $189 | Untapped by current segments "Sustainable Luxury Seekers" — 4.5M users Domain pattern: visit luxury brands + sustainability + eco blogs Personas: Eco-Conscious, High-Net-Worth Avg purchase value: $345 | Premium niche opportunity "Remote Work Parents" — 8.9M users Domain pattern: visit WFH tools + parenting + home office Personas: Parent, Remote Worker, Home Improver Avg purchase value: $134 | Growing post-pandemic segment DISCOVERY METHOD: Unsupervised clustering on domain enrichment features
2
Output: Cohort Revenue Potential

COHORT DISCOVERY — REVENUE POTENTIAL

NEW COHORT REVENUE OPPORTUNITY: ══════════════════════════════════════════════ Cohorts discovered: 23 new high-value segments Total addressable users: 89.4M unique profiles Estimated revenue: $23.4M/quarter from new cohorts Top cohort: "Sustainable Luxury" ($345 avg purchase, 4.5M users) Discovery signals: IAB, Personas, Countries, PageRank patterns

7Geographic Audience Expansion

AI agent expands audience targeting into new geographies by analyzing Countries enrichment data and regional domain patterns — identifying users in new markets who visit domains similar to high-value converters in established markets.

1
Map Audience Quality by Geography
/about Countries IAB Categories Personas OpenPageRank
GEOGRAPHIC AUDIENCE EXPANSION — MARKET MAP ══════════════════════════════════════════════ AUDIENCE QUALITY BY MARKET: US: Profiles: 890M | Enriched: 94% | Avg CVR: 2.8% UK: Profiles: 234M | Enriched: 91% | Avg CVR: 2.4% DE: Profiles: 189M | Enriched: 89% | Avg CVR: 2.1% FR: Profiles: 167M | Enriched: 92% | Avg CVR: 2.3% JP: Profiles: 145M | Enriched: 78% | Avg CVR: 1.9% BR: Profiles: 123M | Enriched: 67% | Avg CVR: 1.4% IN: Profiles: 234M | Enriched: 56% | Avg CVR: 0.9% EXPANSION OPPORTUNITIES: Japan: 78% enriched → target 92% with JP domain intelligence Brazil: 67% enriched → target 85% with BR domain intelligence India: 56% enriched → target 78% with IN domain intelligence
2
Output: Geographic Expansion Plan

GEOGRAPHIC EXPANSION — AUDIENCE PLAN

GEOGRAPHIC EXPANSION PRIORITY: ══════════════════════════════════════════════ 1. Japan — Enrichment gap: 14% | Revenue opportunity: $12.3M/yr 2. Brazil — Enrichment gap: 18% | Revenue opportunity: $8.9M/yr 3. India — Enrichment gap: 22% | Revenue opportunity: $15.6M/yr Total geo expansion revenue: $36.8M/year Method: Add local domain intelligence to enrichment pipeline

8Cross-Device Identity Resolution

AI agent enhances cross-device identity resolution by matching domain visit patterns across devices — using consistent IAB Category and Persona profiles to probabilistically link desktop and mobile browsing sessions for unified audience targeting.

1
Match Domain Patterns Across Devices
/login IAB Categories Personas Countries
CROSS-DEVICE IDENTITY — DOMAIN PATTERN MATCHING ══════════════════════════════════════════════ IDENTITY RESOLUTION METHODS: Deterministic (shared /login): 456M matched pairs (89% confidence) Probabilistic (domain patterns): 234M matched pairs (78% confidence) Domain-enriched matching: 189M additional pairs (72% confidence) DOMAIN PATTERN MATCHING EXAMPLE: Desktop profile: visits nytimes.com, espn.com, bestbuy.com IAB: News, Sports, Shopping | Persona: Active Buyer, Sports Fan Mobile profile: visits nytimes.com (app), ESPN app, bestbuy.com (mobile) IAB: News, Sports, Shopping | Persona: Active Buyer, Sports Fan Match confidence: 89% (same IAB + Persona + geo + timing) CROSS-DEVICE COVERAGE: 67% of Shopper Graph profiles linked
2
Output: Cross-Device Impact

CROSS-DEVICE — RESOLUTION IMPACT

CROSS-DEVICE RESOLUTION RESULTS: ══════════════════════════════════════════════ Matched profile pairs: 879M (deterministic + probabilistic + domain) Cross-device ROAS improvement: +34% Frequency optimization across devices: -45% overexposure Attribution accuracy improvement: +28% Domain-enriched matches added: 189M pairs (27% of total)

9Privacy-Safe Audience Activation

AI agent enables privacy-safe audience activation by creating domain-level audience segments that don't require individual user tracking — using aggregated IAB Categories and Personas at the domain level to target audience types rather than individuals.

1
Build Privacy-Safe Targeting Segments
/about IAB Categories Personas Web Filtering Categories
PRIVACY-SAFE ACTIVATION — DOMAIN-LEVEL TARGETING ══════════════════════════════════════════════ PRIVACY-SAFE SEGMENTS (no individual tracking): "Sports Enthusiast Context" — Any user on 23,456 sports domains IAB: Sports | Personas: Sports Fan | No user ID required Performance: CTR 2.1% | ROAS 5.4x "Luxury Shopper Context" — Any user on 4,567 luxury commerce domains IAB: Shopping > Luxury | Personas: High-Net-Worth | No user ID Performance: CTR 1.8% | ROAS 8.9x "Tech Decision Maker Context" — Any user on 12,345 enterprise tech domains IAB: Technology > Enterprise | Personas: Decision Maker | No user ID Performance: CTR 1.4% | ROAS 6.2x PRIVACY COMPLIANCE: 100% GDPR/CCPA/CPRA compliant No cookies, no user IDs, no PII — pure domain intelligence
2
Output: Privacy-Safe Performance

PRIVACY-SAFE — ACTIVATION REPORT

PRIVACY-SAFE TARGETING RESULTS: ══════════════════════════════════════════════ Privacy-safe segments active: 456 Impressions served without user tracking: 12.3B/month ROAS vs cookie-based targeting: 87% parity (closing gap) Regulatory risk: Zero (no PII, no cookies, no fingerprinting) Revenue from privacy-safe campaigns: $89M/quarter

10Audience Segment Performance Forecasting

AI agent forecasts audience segment performance by analyzing domain intelligence trends — predicting which audience segments will grow or shrink based on changes in domain visit patterns, IAB Category trends, and Persona shifts across the 100M+ domain database.

1
Predict Segment Growth from Domain Trends
/blog /products IAB Categories Personas OpenPageRank
SEGMENT PERFORMANCE FORECAST — Q2 2026 ══════════════════════════════════════════════ GROWING SEGMENTS (increase bids): AI/Tech Enthusiasts: +34% segment growth | Driven by AI domain explosion Sustainability: +23% segment growth | More eco content domains Home Fitness: +18% segment growth | Spring fitness surge STABLE SEGMENTS: General News: +2% segment growth | Mature, stable audience Auto Enthusiasts: +3% segment growth | Consistent interest DECLINING SEGMENTS (reduce spend): Crypto/Web3: -28% segment decline | Domain closures accelerating BNPL Shoppers: -15% segment decline | Regulatory pressure Metaverse/VR: -34% segment decline | Interest waning
2
Output: Segment Forecast Report

AUDIENCE FORECAST — Q2 2026 PREDICTIONS

SEGMENT FORECAST SUMMARY: ══════════════════════════════════════════════ Growing segments: 89 (invest more) Stable segments: 234 (maintain spend) Declining segments: 56 (reduce spend) Revenue from forecast-optimized allocation: +$12.3M/quarter Forecast accuracy (historical): 84% within 10% of actual Signals: IAB Category trends, Persona shifts, PageRank, domain growth
Get in Touch

Interested in AI Agent Domain Intelligence?

For pricing, subscription options, custom database builds, or enterprise partnerships — contact us below.

Power Your AI Agents with Domain Intelligence

Subscribe to the AI Agent Domain Database — continuous access to 100M+ domains, 20 page types each, quarterly refreshes, and real-time change signals.

AI Agent Database View Pricing

Annual subscription includes quarterly data refreshes, change detection alerts, and priority API access.