At this edition of ADC’s Tech Meetup, we zoom in on one of the hardest parts of observational causal inference: unobserved confounders.
Hosted at the ADC office in Amsterdam, this event is for people who have to make decisions from messy data and still care about causality. We’ll look at how Booking.com and ADC use causal inference tools to estimate willingness‑to‑pay and assess robustness to unobserved confounders, and how those methods are turned into pricing and product decisions.
The goal is not just to present methods for situations where randomised experiments aren’t possible, but to show how practitioners actually apply them in industry and to have honest conversations about their robustness and limitations.
Expect two focused talks, plenty of time for questions, and conversations with others facing the same trade‑offs between methods, assumptions, and business pressure.
Please note that this is an in-person event.
Is this meetup for you?
- Data scientists, applied scientists and ML engineers who work with observational data and care about causality.
- Quantitatively minded practitioners in analytics, pricing, and experimentation who need to make decisions when A/B tests aren’t possible or sufficient.
What you will take away
- See how observational causal methods are used in real product and pricing problems, not toy examples.
- Get a look at the tooling and modelling choices of a causal pricing engine.
- Learn how to assess robustness to unobserved confounders and avoid over‑claiming when assumptions are fragile.
- Connect with peers facing the same challenges around dealing with observational data, impossibility of running experiments, and business constraints.
Register for ADC’s Tech Meetup on causal inference
Real cases from product and pricing, with Booking.com and ADC, on handling unobserved confounders in observational data.
From Booking.com and ADC
Hear directly from the industry experts.
Marc Nientker, Operations Lead | ADC
Marc Nientker leads pricing and econometrics work for clients in retail and transportation. He specialises in building causal pricing and markdown engines from observational data, combining robust econometric modelling with scalable machine learning architectures. Marc’s projects focus on turning complex models into fully automated pricing and revenue‑optimisation tools, validated through experiments and embedded in clients’ day‑to‑day decision processes. He regularly works with data science and commercial teams to translate model outputs into actionable pricing, inventory, and promotional strategies that deliver measurable business impact.
Marc's LinkedIn
Lin Jia, Senior Data Scientist | Booking.com
Lin Jia is a Senior Data Scientist at Booking.com, working on experimentation, observational causal inference, and applied machine learning for data-driven product decisions. Her work focuses on designing rigorous measurement approaches using both randomized experiments and observational methods, and applying causal machine learning techniques such as DoubleML in real-world product settings. She is also involved in integrating generative AI tooling into experimentation platforms to support analysis workflows and knowledge retrieval. Her work on sensitivity analysis for causal machine learning has been presented at CausalML Workshop at KDD and the Causal Data Science Meeting.
Lin Jia's LinkedInAbstract
Willingness-to-pay: A modern causal inference approach
In many markets, organizations are moving from coarse, rule-based pricing structures toward more flexible, continuously adjusted prices that better reflect differences in customer willingness-to-pay. A central challenge in this transition is estimating the underlying demand curve: understanding how demand responds to price changes across products, moments, and customer contexts. This is fundamentally a causal problem rather than a predictive one, because observed prices are influenced by many factors that also affect demand, creating confounding influences that obscure true price sensitivity.
In this talk, I present an approach to estimating willingness-to-pay curves in the presence of such unobserved confounders, building on recent advances in econometrics. The key insight is that many hidden factors influence groups of related products or contexts simultaneously, though with varying intensity. By modelling this shared structure directly, the method enables robust demand estimation without requiring explicit knowledge of the confounding processes, providing a practical and scalable foundation for more accurate pricing decisions in complex environments.
Sensitivity analysis in causal ML: When to trust observational estimates
Observational causal inference inevitably relies on assumptions about unobserved confounding. Sensitivity analysis provides a structured way to evaluate how robust our conclusions are to potential violations of these assumptions. In this talk, I will discuss how sensitivity analysis can be applied in practice using Double Machine Learning, drawing on a real-world use case from Booking.com. Beyond the methodology itself, I will focus on practical questions practitioners face: how to benchmark confounding strength, how to interpret robustness metrics, and common pitfalls when sensitivity analysis is used to justify weak causal designs.
Programme
17:00 – 17:30 | Door opens
17:40 – 17:50 | Introductory remarks
17:50 – 18:30 | Willingness-to-pay: A modern causal inference approach
18:30 – 19:10 | Sensitivity Analysis in Causal ML: When to Trust Observational Estimates
19:10 onwards | Walking food, networking discussions
20:00 | Closing
Event details
- Date and time: 21 April | 17:30–20:00
- Location: ADC Consulting, Amsterdam office, De Ruijterkade 7, 1013 AA Amsterdam
Get in touch
Questions about the event? Contact Luisa Carrer at luisa.c@adc-consulting.com