A Conversation with Luca Fiaschi of PyMC Labs: Synthetic Consumers & the Future of Product Testing
How Colgate, Bayesian statistics, and AI-generated “digital clones” are reshaping how brands develop, test, and optimize products.
Before joining PyMC Labs, Luca Fiaschi spent his career blending cutting-edge research with real-world execution. He built early data teams at HelloFresh, helped scale Rocket Internet ventures across the world (including Lazada, later part of Alibaba), and worked across everything from forecasting and churn modeling to ad-tech bidding engines and experimentation systems.
With a PhD from Heidelberg focused on machine learning for high-throughput microscopy, Luca is uniquely positioned at the intersection of Bayesian statistics, generative AI, and practical business problems. Today, he leads the generative AI vertical at PyMC Labs, a company built on the open-source PyMC library, helping enterprises solve complex decision workflows using statistical reasoning and AI agents.
Listen/Watch Our Interview:
In this episode of Retailgentic, Scot sits down with Luca to unpack a research paper that blends academic rigor with real-world implications:
“LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings.”
The paper, co-authored with Colgate-Palmolive researchers, explores whether AI can accurately simulate human reactions to product concepts, enough to replace or accelerate traditional consumer panels, which are slow, expensive, and hard to scale.
This one goes deep, but in ways that any retail or AI leader should care about. Scot and Luca discuss:
Colgate’s Challenge: How to test product concepts faster and at scale.
Synthetic Consumers: AI models that react to products like human panels.
Accuracy Breakthrough: Reaching ~73–74% agreement with real consumers.
Fixing LLM Failure Modes: Why naive prompts don’t work, and what does.
Bayesian Reasoning: Adding uncertainty so AI stops being confidently wrong.
Smarter A/B Testing: Using AI to pre-screen ideas before running live experiments.
Digital Clones: Future consumers earning money by sharing preference data safely.
Simulated Populations: Matching real audiences for testing and predictions.
AI isn’t just helping brands write copy or generate images, it’s beginning to think like their customers. If synthetic consumers continue to evolve at this pace, product development, A/B testing, and personalization may look completely different in just a few years.
🎧 Listen wherever you get your podcasts: Spotify, Apple, etc.
Or, we have a video version on YouTube (make sure to subscribe!) →
Loving this Convo? We’re just getting started.
Stay on top of what’s happening in agentic commerce, from the tools powering it to the founders building with it by following us across platforms:
Podcast: https://retailgentic.transistor.fm
Instagram: https://www.instagram.com/retailgentic/
We're here to make sense of what’s next in retail + AI and bring you along for the ride. Subscribe and stay tuned!
Really smart work here. The 73% agreement with real panels is impressive, but the Bayesian uncertainty layer is what makes it actually usable for product decisions. Most synthtic consumer models collapse when concepts shift from familiar to novel categories where training data thins out.This apprach at least signals when it's extrapolating vs interpolating. Be curious to see how calibration holds across different product verticals.