AANZ August Expert Session | with Jeffrey Paine Golden Gate Ventures

This podcast link and below summary is of Jeffrey Paine’s Expert Session held online August 2024. 

Podcast: Reinventing Venture Capital with AI: Insights from Jeffrey Paine of Golden Gate Ventures 

This document provides a comprehensive executive summary of an Angel Association New Zealand expert session led by Jeffrey Paine, the founding partner of Golden Gate Ventures, a prominent venture capital firm based in Singapore. Paine, a respected leader in the Asian startup ecosystem with experience spanning Silicon Valley, Dubai, and more expansive Asia, shared his insights on using Artificial Intelligence (AI) for portfolio management, tracking, monitoring, and preempting market trends. The session delved into Golden Gate Ventures’ data-driven approach, their successful prediction of recent market downturns, and their vision for the future of AI-powered venture capital. 

A Veteran’s Perspective on a Changing Landscape 

Jeffrey Paine, who has been involved with the New Zealand startup scene for over 15 years, emphasised that successful companies are not formed in isolation. He highlighted the crucial role of seed funding, experience, mentors, and strong networks for entrepreneurs. Angel Association New Zealand aims to equip early-stage investors with the necessary tools, resources, networks, and support to make informed investment decisions, ultimately fostering the growth of exceptional startups. 

Paine then took the audience through a chronological account of Golden Gate Ventures’ journey since 2019, which has been marked by significant shifts in the Southeast Asian venture capital landscape. He noted that 2018 saw a massive influx of new funds into the region, leading to inflated startup valuations, particularly in Vietnam. Recognising this unsustainable trend, Golden Gate Ventures publicly voiced its concerns in early 2019 and began to rethink its investment strategy to gain a competitive edge in an increasingly crowded market. 

The Power of Data: Predicting Market Crashes 

The turning point for Golden Gate Ventures came through a deep dive into data analysis. Paine revealed that they stopped investing in April 2021, a move met with scepticism by their team and portfolio founders, who were experiencing what seemed like the best year ever with abundant capital and rapid hiring. However, this data-driven decision allowed them to predict the market crash in 2022, being only two months off due to the unforeseen impact of the war. Consequently, they began selling assets from their earlier funds in 2021 and advised their portfolio companies to prepare for a downturn. 

Their predictive capabilities extended further as they correctly anticipated the more severe crash of 2023 and the subsequent wave of mergers and shutdowns in 2024. By leveraging data, Golden Gate Ventures avoided investing in failed business models, such as quick commerce and Buy Now Pay Later (BNPL). This proactive approach has significantly improved their hit rates (10X or more returns) from Fund 3 and 4, with failure rates also dropping considerably. Notably, between April 2021 and the time of the session, they had only invested in nine companies, all at the seed stage, a testament to their highly selective and data-informed strategy. 

Fund Sizing, Market Focus, and the Elusive Home Run 

Paine emphasised the critical importance of fund sizing to ensure the potential for a 4X return. This involves extensive data crunching to understand the number of companies being started, the backgrounds of founders, and geographical nuances. He highlighted a stark comparison between Southeast Asia and the US, noting that Southeast Asia averages only 0.3 unicorns per year, contrasting sharply with the 100s of unicorns in the US. This reality dictates that a fund might only identify one to two potential winners in Southeast Asia every four-year investment period. 

This understanding has fundamentally shifted Golden Gate Ventures’ investment strategy and criteria. Their number one and two criteria are now market size and speed to market, respectively, taking precedence over the team at the initial stage. Paine explained that a large market that takes too long to penetrate will still lead to losses. He provided a simple yet effective metric for assessing market potential: achieving $100 million in Annual Recurring Revenue (ARR) by year eight from the first institutional funding round. This benchmark helps quickly determine if a startup has the potential to generate the magnitude of returns required for venture capital. 

Paine underscored that in venture capital, home runs matter most; strikeouts are inevitable. They focus on selection, which he believes constitutes 70% of successful VC investing, with portfolio management making up the remainder. He argued that the best companies often require minimal hands-on management, making the initial selection the most critical factor. He also touched upon the concept of power law distribution, noting its variations outside of key US tech hubs like the West Coast, impacting fund life, size, and geographical investment strategies. 

Leveraging Data Science and AI: Predicting the Future 

Paine discussed the inherent inefficiencies in venture capital, often driven by human biases related to founder demographics and backgrounds. He also acknowledged the sobering reality that many startups still fail even with rigorous due diligence. To mitigate these challenges and improve selection, Golden Gate Ventures heavily relies on data science to increase its deal flow visibility, aiming to see close to 100% of deals in its target markets. He emphasised the need to know the future, predicting market size and timing, and cautioned against solely relying on founders’ projections, especially after the exuberance of 2021, where due diligence often took a backseat. Building a strong network remains essential for deal flow and reference checks, but Paine highlighted the increasing necessity of using data to augment and enhance these subjective aspects. 

He revealed a growing trend of VC funds using data, with approximately 70 funds in the West (US and Europe) and a few in Asia (mainly India) adopting this approach. Of these, only a smaller subset (around 7-8 in the West) truly leverage AI, while the rest primarily use basic data analytics. However, Paine predicts a significant increase in AI adoption within the VC industry in the next two to three years. He noted that more considerable funds (Series B & C) tend to adopt AI more readily due to higher management fees enabling them to hire data scientists and the availability of more substantial data to analyse. 

Golden Gate Ventures’ AI-driven process involves two key stages: reasoning and prediction. 

• Reasoning: This involves using data to crunch trends and conduct research on market size and timing. Paine mentioned that what previously took his team two weeks to generate now takes him 20 seconds using their AI tools. He analyses around 1400 sectors every two months, identifying changing signals and informing their internal research process, which involves a more manual deep dive. 

• Prediction: This stage employs machine learning models to analyse data from companies that have succeeded, incorporating factors like team dynamics (founder backgrounds, age, education, diversity, early hires), funding history (Brownian motion measuring capital momentum), and signal data (online mentions, app store rankings, GitHub activity). 

Paine stated that Golden Gate Ventures has already scored 94% of all Southeast Asian companies, allowing them to quickly identify and often preemptively decline investment opportunities. Their strategy has shifted towards a 70% outbound approach, where they conduct in-depth research on specific sectors for at least six months before proactively reaching out to promising founders who may not even be actively fundraising. This deep understanding often positions them as more knowledgeable than the founders, leading to different and more insightful conversations. Their outbound strategy has been successful in securing deals even at lower valuations. 

AI for Portfolio Management and the Future of VC 

Beyond deal sourcing and selection, Golden Gate Ventures utilises data and AI for portfolio management. By continuously monitoring portfolio company data, they can inform founders about market changes, make better follow-on investment decisions, and determine when to cut losses or pursue exits. Paine shared an anecdote about their early exit predictions, which were often met with disbelief by founders and boards who couldn’t understand their rationale at the time. 

Looking to the future, Paine envisions a completely automated VC fund within the next three years, potentially spearheaded by himself or someone else in the industry. He highlighted their ongoing efforts in areas like: 

• Personality Data Analysis: Leveraging data from platforms like LinkedIn to understand founder personalities and their correlation with success. 

• Automated Market Sizing and Timing Research: Further automating the deep sector analysis to achieve near real-time insights. 

• Deep Tech Prediction: Developing models to predict promising deep tech spin-outs from universities. 

• AI-driven Networking: Using graph neural networks to identify key individuals (other VCs, angels, professors) to network with for improved deal access and efficiency, especially during travel. 

• Automation of Internal Workflows: Utilising Large Language Models (LLMs) to automate tasks like memo generation and investor introductions, aiming to connect portfolio companies with the most relevant follow-on investors based on data-driven probabilities. 

Paine also shared some key tips for angel investors and VCs: 

• Avoid asking investors to refer to other investors if you are not investing yourself. 

• Instead, ask an active portfolio company of an investor for an introduction. 

• Specialise in a sector to develop deep expertise and accelerate due diligence. 

• Understand that angels prioritise founders, while VCs often focus on the team. 

• Recognise the lack of communication between seed investors/angels and later-stage funds and try to bridge this gap. 

• Thoroughly understand the dynamics of your local ecosystem, including the power law, the number of new companies, booming sectors, and talent clusters. 

• Work backward from desired returns (e.g., 4X for a fund) to inform investment decisions, similar to how founders should work backward to achieve their market size goals. 

• For New Zealand investors, Paine strongly advised considering investments in the US to build networks and facilitate the crucial later-stage funding rounds for promising New Zealand companies, as US money often dominates the final rounds of unicorns. He suggested rethinking government matching fund criteria to encourage more international, particularly US-focused, seed-level investments. 

New Zealand’s Potential and Small Fund Advantages 

Addressing specific questions about trends in New Zealand, Paine acknowledged the strength in B2B enterprise SaaS, Agritech, and SpaceTech. However, he emphasised that for New Zealand companies to succeed globally and attract foreign investment, they must aim to be in the top four globally in their chosen niche, not just the best in New Zealand or even developed Asia. 

Paine offered a positive perspective regarding small funds (less than NZD 50 million). He argued that smaller funds can often achieve the targeted 4X returns more quickly due to their inherent flexibility and the dynamics of return modelling. While LPs will scrutinise their ability to attract quality deals and manage the fund with the available fees, a well-defined strategy, explicit deal sourcing advantage, and a realistic portfolio size can make smaller funds highly attractive and potentially more profitable than more considerable funds, especially in the current market environment. 

Diving Deeper into the AI Toolkit 

Paine offered insights into their AI toolkit for attendees interested in the technical aspects. He highlighted that the most challenging part of machine learning in this context is data: acquiring it, cleaning it, wrangling it, and performing feature engineering. They are leveraging Large Language Models (LLMs) to enhance their data, filling in missing information and correcting inaccuracies. While commercial solutions exist, Paine found them lacking and suggested that building a simple in-house solution while studying academic papers in the field (identifying around 5-6 perfect ones out of about 30) can be a more practical approach. He also recommended engaging AI and data science interns to assist with data acquisition and experimentation. In terms of specific machine learning models, he mentioned that Random Forest works, but his personal favourites are XG Boost and Cat Boost. He encouraged attendees to reach out for guidance on learning these techniques. 

Finally, Paine noted the emerging trend of AI companies, predicting a significant influx in New Zealand and globally in the coming year and a half. He advised investors to be cautious yet recognise the opportunities within this space. 

In conclusion, Jeffrey Paine’s session provided a compelling look into the data-driven future of venture capital. His insights, grounded in real-world experience and a willingness to share the intricacies of Golden Gate Ventures’ AI-powered strategies, offered valuable lessons for angel investors and venture capitalists in New Zealand and beyond. His emphasis on data-informed decision-making, global market focus, and the potential of AI to transform every stage of the investment process painted a vivid picture of the evolving landscape of startup funding.