Episode

Eoin O'Mahony on Why Good Metrics Still Lead to Bad Decisions — and How to Fix It

Eoin Mahony—data science partner at Lightspeed, former Uber science lead, and co-designer of the system that kept NYC’s Citi Bikes available across the city—argues that positive metrics are meaningless if you don’t understand the mechanism behind them. At Uber, he often blocked launches that looked good on paper but didn’t make sense in practice. Now in venture, he’s applying that same rigor to unstructured data—using GenAI to scale a kind of work that’s long resisted systematization.

April 23, 2025

Listen on

Guest

Eoin O'Mahony

Lightspeed Ventures

Eoin O'Mahony is Partner, Data Science at Lightspeed, where he’s focused on building a strong internal data engine to support data-driven decision-making across the firm. He spent eight years working in applied science at Uber, where he built and led high-performing science teams. Before that, at Citi Bike, he used applied analytics, machine learning, and optimization to solve real-world transportation problems in New York City. Eoin holds a PhD and M.S. in Computer Science from Cornell University and a BSc in Computer Science from University College Cork. His technical background is helping make data a core pillar of how Lightspeed operates. Originally from a small town in the Irish countryside—where everyone knew their neighbors—Eoin brings that same sense of warmth, community, and willingness to help to the tight-knit world of venture capital. “The broader VC industry is going through an inflection point in terms of how it leverages data,” Eoin says. “I want Lightspeed to be setting the standard and shaping the direction the industry is taking.” Outside of work, Eoin enjoys exploring new restaurants with his wife and two young kids. He also loves reading and cooking—two hobbies he says sharpen his execution mindset: “When something’s worth doing, it’s worth doing well.”

HOST

Hugo Bowne-Anderson

Delphina

Hugo Bowne-Anderson is an independent data and AI consultant with extensive experience in the tech industry. He is the host of the industry podcast Vanishing Gradients, a podcast exploring developments in data science and AI. Previously, Hugo served as Head of Developer Relations at Outerbounds and held roles at Coiled and DataCamp, where his work in data science education reached over 3 million learners. He has taught at Yale University, Cold Spring Harbor Laboratory, and conferences like SciPy and PyCon, and is a passionate advocate for democratizing data skills and open-source tools.

Key Quotes

Key Takeaways

1. Positive Metrics Can Mislead — Mechanism Matters More

Eoin blocked product launches at Uber when teams couldn’t explain why a metric had improved. Without a clear causal explanation, those “wins” often turned out to be losses.

2. Early-Stage Products Don’t Need A/B Tests — They Need Impact

For new products, Eoin didn’t bother with small-effect experiments. If a change didn’t create a visible spike in time series data, it wasn’t worth pursuing.

3. Simple Algorithms Win in the Real World

At Citi Bike, Eoin scrapped complex optimization for plain, actionable routing instructions that teams could follow. He optimized for execution, not elegance.

4. Complex Systems Break Simple Assumptions

Marketplace dynamics often invalidate basic experimentation. Eoin shows how network effects, time lags, and interactions can make even clean A/B tests misleading.

5. GenAI Might Finally Scale Judgment Work in VC

At Lightspeed, Eoin is building tooling to extract signal from unstructured data — applying scientific rigor in a domain that’s long resisted systematization.

‍

You can read the full transcript here.

‍

00:00 Positive Metrics Can Mislead — Mechanism Matters More

03:18 Eoin's Journey with Citi Bike05:56 Challenges and Solutions at Citi Bike

14:05 Uber: From Two Wheels to Four

19:47 Navigating the COVID-19 Crisis at Uber

26:21 Transition to Venture Capital at Lightspeed

28:05 Joining Lightspeed: A New Opportunity

29:19 Exciting Projects at Lightspeed

30:08 Exploring Generative AI Tools

31:27 The Rapid Evolution of AI