PitchBook's new tool uses AI to predict which startups will successfully exit | TechCrunch

Abstraction: ML startup exit predictor accuracy claims and embedded bias risks

Key points:

• PitchBook VC Exit Predictor generates probability score for acquisition, IPO, or failure; requires at least two prior VC financing rounds; back-tested at 74% accuracy on historical exits (Blockchain.com, Revolut, Bitso)
• Gartner predicts 75%+ of VC executive reviews will use AI/data analytics by 2025
• Tool held favorable outlook on crypto companies despite sector-wide decline — slow to adjust to market shifts; vulnerable to black swan events like pandemics
• Harvard Business Review 2020 experiment: investment algorithm picked white, male-founded startups due to historical funding disparities in training data
• CB Insights' Mosaic: 4 of 6 disclosed "signals" are proxies for race, socioeconomic status, gender, and disability
• PitchBook itself found a 1% difference in success predictions between male and female CEOs despite claiming gender-blindness

Connections: Pitchbook · Cb Insights · Machine Learning · Predictive Analytics · Algorithmic Bias

Source: https://techcrunch.com/2023/03/20/pitchbooks-new-tool-uses-ai-to-predict-which-startups-will-successfully-exit/