PHBench is an open benchmark that predicts which Product Hunt launches will raise Series A funding. It analyzes 67,292 launches over seven years, using machine learning models to identify key signals like daily rank and upvotes. The best model achieves 4.7x lift over random, helping startups and investors spot future winners.
Free
How to use PHBench?
PHBench enables users to submit predictions for new Product Hunt launches and receive weekly forecasts on Series A potential. It solves the problem of identifying high-potential startups early, leveraging a leaderboard of models trained on historical data. Users can filter by model type, explore signal importance, and cite the benchmark in research.
PHBench 's Core Features
Open benchmark predicting Series A funding from Product Hunt launch signals, trained on 67,292 launches over seven years for robust analysis.
Leaderboard ranking models by F0.5 score, AP, REC, and AUC, with top ensembles achieving 0.284 F0.5 and 4.7x lift over random.
Identifies 12 key predictive signals, including daily rank (3.5x lift) and maker follower count, while filtering out noise like raw upvote count.
Reproducible methodology with manually audited labels, documented features, and a hash-pinned test set for fair evaluation.
Weekly predictions and submission system for new launches, with integration of LLM-based models like Gemini 3 Flash for zero-shot analysis.
PHBench 's Use Cases
Startup founders use PHBench to gauge Series A potential of their Product Hunt launch, optimizing timing and engagement.
Venture capitalists leverage the benchmark to filter high-probability startups from thousands of launches, saving time.
Data scientists train custom models on the open dataset, improving prediction accuracy for investment strategies.
Researchers cite PHBench in academic papers to validate Series A prediction methods and benchmark new algorithms.
Product Hunt makers analyze signal importance to boost launch performance, focusing on daily rank and upvote interactions.