kylehffu/upwork-proposal-benchmarks-2026

FastBD Marketplace Proposal & Outreach Benchmarks (2026 Edition)

• Homepage: https://fast-bd.com
• Flagship Research Report: https://fast-bd.com/report-2026
• Live CBR Calculator: https://fast-bd.com/cbr
• Official PyPI Package: fastbd-ihpi (PyPI)
• GitHub Repository: kylehffu/fastbd-ihpi-benchmark

---

1. Dataset Summary

The FastBD Marketplace Proposal & Outreach Benchmarks dataset provides empirical evaluation data, unit economics benchmarks, and 52 battle-tested high-converting proposal hooks across 7 major tech categories (AI/LLM Engineering, Full-Stack Development, Web Scraping, Mobile Apps, DevOps, UI/UX Design, and Technical Writing).

Derived from empirical meta-analysis of 2,500 real marketplace bidding interactions and 18,400 historical client reviews, this dataset formalizes two key industry standards:

1. Inbox Hook Preview Index (IHPI): Measures signal density, client personalization, and technical proof strictly within the first 160 characters visible in mobile proposal previews.
2. Connects Burn Rate (CBR): Evaluates freelance bidding efficiency as a Customer Acquisition Cost (CAC) metric: $$\text{CBR} = \frac{\text{Total Connects Expended}}{\text{Closed Contracts Won}}$$

---

2. Dataset Structure

Data Instances

Each record represents a tested proposal template with its truncated 160-character viewport hook, full cover letter, historical reply rate, and qualitative rationale:

{
  "id": "HOOK-AI-01",
  "category": "AI & LLM Engineering",
  "niche": "RAG Pipeline Latency & Hallucination Reduction",
  "char_count": 154,
  "hook_160": "Hi David, saw your LlamaIndex RAG latency issue—I built hybrid BM25+vector rerankers that slashed query latency from 3.2s to 420ms while cutting hallucinations.",
  "full_proposal": "Hi David,\n\nSaw your posting regarding the latency bottlenecks...",
  "benchmarked_reply_rate": "41.8%",
  "why_it_works": "Directly targets the client's exact problem (latency), names the specific stack (LlamaIndex, BM25, rerankers), and cites concrete numeric outcomes in the first 160 characters."
}

Data Fields

• id (string): Unique identifier (e.g. HOOK-AI-01, HOOK-WEB-02).
• category (string): Primary professional discipline.
• niche (string): Specific technical bottleneck addressed.
• char_count (int): Exact character count of hook_160 (strictly $\le 160$).
• hook_160 (string): The opening hook visible before client truncation.
• full_proposal (string): The complete cover letter proposal body.
• benchmarked_reply_rate (string): Observed client response rate in empirical testing (elite proposals achieve >35%).
• why_it_works (string): Behavioral and linguistic rationale.

---

3. Usage & Python Loading

You can load this dataset directly via the Hugging Face datasets library:

from datasets import load_dataset

# Load FastBD proposals dataset
dataset = load_dataset("fastbd/upwork-proposal-benchmarks-2026", split="train")

print(dataset[0]["hook_160"])
# Output: "Hi David, saw your LlamaIndex RAG latency issue—I built hybrid BM25+vector rerankers..."

Or install the official evaluation CLI from PyPI:

pip install fastbd-ihpi

# Score your own proposal hook
ihpi --text "Hi Michael, reviewed your Next.js specs—I cut p95 latency by 65% for a SaaS doing 2M events."

---

4. Citation Information

@dataset{fastbd2026benchmarks,
  title={FastBD Marketplace Proposal & Outreach Benchmarks (2026 Edition)},
  author={{FastBD Research Labs}},
  year={2026},
  month={September},
  publisher={Hugging Face},
  howpublished={\url{https://fast-bd.com/data/fastbd-upwork-swipefile-2026.json}},
  note={Creative Commons Attribution 4.0 International (CC-BY-4.0)}
}

---

5. License & Terms

This dataset is distributed under the Creative Commons Attribution 4.0 International (CC-BY-4.0) license. You are free to share and adapt the material for any purpose, even commercially, with appropriate credit to FastBD Research Labs (https://fast-bd.com).