PersonaGen-Enterprise: B2B Buying Intelligence Dataset

5,000 enterprise buyer personas. 47,019 real search queries. 200 brand matchups across 15 industries.

Rankfor.AI is releasing PersonaGen-Enterprise on HuggingFace: a structured dataset of B2B buying committee intelligence built to help researchers, marketers, and AI teams understand how enterprise buyers actually discover, evaluate, and choose technology vendors.

This follows our PersonaGen-15K release (consumer personas) and accompanies the PersonaGen-149K paper submitted to Springer The Journal of Supercomputing. The enterprise dataset is the first public resource to model full buying committee dynamics at scale, grounded in real search behavior rather than synthetic assumptions.

---

What's Inside

Dimension	Count
Enterprise B2B buyer personas	5,000 (anonymized from 130K)
Real search queries (Perplexity AI + Reddit)	47,019
Competitive brand queries	7,500
Industries covered	15
Buying committee roles	42
Brand matchups	200
LLMs used for generation	5

Modeled industries include: FinTech/Banking, Healthcare/Pharma, E-commerce, Insurance, SaaS/MarTech, FMCG, Energy/Utilities, Manufacturing, Advertising Agencies, Professional Services, Automotive, Media/Entertainment, Real Estate, QSR/Food, Mobility/Travel.

Buying committee roles span: CFO, CTO, CMO, CISO, VP of IT Operations, Head of Digital Banking, Director of Procurement, Enterprise IT Manager, and 34 more enterprise-specific roles.

LLMs used: Gemini 3 Flash, GPT-5.2, Claude Haiku 4.5, Claude Sonnet 4.6, Grok 4.1. Each persona was validated for cross-model agreement before inclusion.

---

How It's Different

PersonaGen-Enterprise was built around a single question: what does a real buying committee actually look like, and what does each member search for?

Grounded in real queries. The 47,019 search queries were sourced from Perplexity AI and Reddit, reflecting actual language enterprise buyers use when they are mid-evaluation. These are real queries people typed, tagged with industry, buying stage, brands mentioned, and intent.

Full buying committee modeling. Each persona carries structured attributes across the entire purchase lifecycle: buying stage, budget authority level, evaluation criteria, deal-breakers, and live objections. A CFO persona at the approval stage looks very different from a security architect at the technical evaluation stage. The dataset captures both.

Competitive brand intelligence. The 200 brand matchups reflect real co-occurrence patterns in buyer research. Databricks vs Snowflake appears in 1,622 queries; HubSpot vs Marketo in 1,162; Power BI vs Tableau in 1,271. These are not curated pairs: they emerge from how buyers actually frame comparisons when searching.

Multi-model generation and validation. The 130K full dataset was generated by 5 different LLMs. Cross-model agreement ranges from 0.72 to 0.88 cosine similarity across industries. Personas where models diverged significantly were flagged. What remains reflects stable, convergent buying intelligence.

---

Quick Start

from datasets import load_dataset
import json

# Load enterprise personas (default config)
ds = load_dataset("rankfor/PersonaGen-Enterprise")
personas = ds["train"]

print(f"Total personas: {len(personas)}")

# Inspect a persona
p = personas[0]
print(f"Industry: {p['industry']}")
print(f"Role: {p['buying_committee_role']}")
print(f"Buying stage: {p['buying_stage']}")
print(f"Budget authority: {p['budget_authority']}")

# Parse search queries
queries = json.loads(p['typical_queries'])
print(f"Typical queries: {queries[:3]}")

# Load real search queries
real_q = load_dataset("rankfor/PersonaGen-Enterprise", "real_queries")
print(f"Real queries: {len(real_q['train'])}")

# Load competitive query banks
banks = load_dataset("rankfor/PersonaGen-Enterprise", "query_banks")
print(f"Query bank entries: {len(banks['train'])}")

---

Research Findings

The full PersonaGen-1M study surfaces several patterns relevant to anyone working in B2B go-to-market or AI visibility:

1. Healthcare/Pharma has the highest Evaluating-stage concentration (26%). This signals the most active enterprise buying intent across all 15 industries. If you sell to healthcare, the buyers are further down the funnel than in any other vertical.

2. Integration complexity is the #1 deal blocker. 33.2% of all objections cluster around integration concerns. Vendors who lead with integration stories in their content have a structural advantage across industries.

3. 99.8% of enterprise queries carry commercial intent. Compared to 18% in the consumer PersonaGen-15K dataset. Enterprise buyers search differently: they are comparing vendors, not browsing.

4. Budget authority does not always match seniority. Directors may be Approvers at mid-market companies but only Influencers at enterprise scale. The same title means different purchasing power depending on company size.

5. The most-contested vendor matchups are cross-category. Databricks vs Snowflake (1,622 co-occurrences), Power BI vs Tableau (1,271), HubSpot vs Marketo (1,162). Buyers do not compare within narrow categories; they compare across the tools they actually use together.

---

Full Dataset Access

The 5,000-persona public release is available on HuggingFace under CC BY 4.0.

The complete 130,000-persona dataset includes:

• All buying committee roles (42 unique)
• Complete vendor stack details (not redacted)
• All 5 LLM generation models
• Full content map (3,104 role-to-content mappings)
• Deduped embeddings for semantic search

Available through Rankfor.AI enterprise data licensing. Contact: sales@rankfor.ai or visit rankfor.ai.

---

Links

• HuggingFace dataset: rankfor/PersonaGen-Enterprise
• Consumer personas: rankfor/PersonaGen-15K
• GitHub: Rankfor/rankfor-open
• Paper: PersonaGen-1M

---

Citation

@dataset{rankfor2026personagen_enterprise,
  title     = {PersonaGen-Enterprise: B2B Buying Intelligence Dataset},
  author    = {\.{Z}atuchin, Dmitrij},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/datasets/rankfor/PersonaGen-Enterprise},
  license   = {CC BY 4.0}
}

---

Dataset license: CC BY 4.0. Free to use, share, and adapt with attribution.

Published by Rankfor.AI: AI Visibility Intelligence for B2B brands.