cAI24 London: the Causal AI Conference by causaLens

Watch here

• 
  Darko Matovski Founding CEO, causaLens

Be inspired about the business rationale of how Europe’s largest Outdoor eCommerce Retailer Bergfreunde.de (known in the UK as www.alpinetrek.co.uk/) uses Causal AI to intervene on key operational business metrics and forecast each week’s sales and profit and learn how to decode eCommerce operations algorithmically

• 
  Christopher Barth Head of Data & Analytics, Bergfreunde

While experimentation is the golden standard for causal inference and is widely adopted in the industry, it is sometimes infeasible or undesirable. In these settings, a common causal estimate is the incremental impact of a feature or program that is released to the whole customer base, but only a subset of users adopts it or subscribes to it. In this talk, we present some practical learnings from the fashion industry, with an application on the incrementality of subscription programs.

• 
  Matteo Courthoud Senior Applied Scientist, Zalando

Promotions are crucial in e-commerce for increasing user engagement and building customer loyalty. These incentives naturally raise causal questions, as they function as interventions that alter the business’s status quo. Effectively managing the cost-effectiveness of promotions is crucial for creating campaigns that benefit both customers and businesses. This involves addressing a dual causal inference challenge: balancing the uplift in rewards, such as conversion and revenue, against the corresponding increase in costs. In e-commerce, promotional costs and rewards are typically incurred only upon conversion, resulting in zero cost and reward for non-conversions. Given that conversion rates are often below 10%, the data is predominantly zero-inflated, with over 90% of the observations providing limited information about the targets of interest. In this presentation, we demonstrate that in scenarios with triggered costs and rewards, effective promotion modeling can be achieved by applying uplift modeling techniques exclusively to converted data from randomized experiments (e.g., A/B tests). This approach enables the accurate calculation of key metrics, such as lift, relative metrics (revenue, utility, etc.), and Return on Investment (ROI), while also reducing training times by more than 90% without compromising model performance.

• 
  Hugo Proença Senior Machine Learning Scientist, Booking.com

At Croud, we are taking media planning and optimization to the next level by combining causal AI with Media Mix Modeling (MMM). This integration not only enhances our models’ ability to explain and analyze marketing outcomes but also makes them more robust and insightful. By bringing together the best of both approaches, we are able to offer our clients deeper insights and more effective strategies. In this presentation, we will dive into how this blend of techniques is transforming the way we approach media planning.

• 
  Fernanda Hinze Data Scientist, Croud

• 
  Dr Athanasios Vlontzos Research Scientist, Spotify

In this presentation, I’ll explore the central role of Causal Inference in credit allocation and optimization within the banking business. I’ll go over how determining the amount of credit credit and setting the right interest rate can be framed as a Conditional Average Treatment Effect estimation problem. Since both credit and price are continuous treatments, I’ll talk about the main challenges that come with them and some exciting attempts at solving those problems.

• 
  Matheus Facure Staff Data Scientist, Nubank

Causal reasoning is vital for effective reasoning in many domains, from healthcare to economics. In medical diagnosis, for example, a doctor aims to explain a patient’s symptoms by determining the diseases causing them. This is because causal relations, unlike correlations, allow one to reason about the consequences of possible treatments and to answer counterfactual queries. In this talk I will present some recent work done with my collaborators about how one can learn and reason with counterfactual distributions, and why this is importantly for decision making. In all cases I will strive to motivate and contextualise the results with real word examples.

• 
  Ciarán M. Gilligan-Lee Head of Causal Inference Lab & Senior Research Manager, Spotify

Causal AI is already used in the energy sector, but it’s still key for advancing our operations. We use it to quickly turn data into insights and then into actions, which is crucial for managing large operations. We’re looking at how Causal AI is currently used in different parts of the energy industry and how it might be used in the future. We also consider how today’s causal methods are setting the stage for new technologies, including a look at an advanced wind farm concept.

• 
  Cristiano Da Cruz Machine Learning Researcher, BP
• 
  Robin Yellow Principal, Digital Science & Engineering, BP

The dynamic retail environment presents unique challenges for causal inference in optimal pricing. Evolving business tactics, seasonality, changing product assortments, and censored demand create highly contextual, heterogeneous, and time-varying price response effects. In this presentation, we will demonstrate how causal ML is utilized for markdown optimization at Marks & Spencer. We will explain our approach to predicting price elasticities, accounting for uncertainty, and address the pitfalls of using aggregated data for causal inference. Our experiences illustrate how these methods can lead to more robust and effective decision-making processes in complex retail environments.

• 
  Mindaugas Zickus Lead Data Scientist, Marks and Spencer

• 
  Gideon Wilkins Group Head of Research, McCann Worldgroup

When we think about causal modeling, we often focus on (critically important) assumptions necessary to make valid causal inferences. The value of this perspective is hard to overestimate. At the same time, the true goal of applied causal inference is to improve our decision-making rather than just meet causal assumptions. In this talk, we will take a look at a set of methods that allow us to make causal inferences under violated assumptions. Although some, if not all, of these methods might not lead to inferences as precise as we’d like, they can help us solve our real problem – making better decisions.

• 
  Aleksander Molak Causal ML Tutor / Causal Bandits Host, University of Oxford / Causalpython.io

In retail banking, understanding the causal relationships between variables is crucial for making informed decisions that enhance customer experience and optimize financial outcomes. At BBVA AI Factory, causal inference techniques are improving our approach to complex challenges where traditional A/B testing is infeasible or removing bias becomes a hard problem. Also, it has significantly enhanced financial health metrics and risk management processes. By utilizing causal inference, we can find out the true impact of various interventions on customer behavior and financial stability, leading to more accurate risk assessments and tailored financial advice. This approach addresses the limitations of conventional experimental methods, which can be impractical or ethically challenging, by providing robust alternative solutions. This presentation will showcase three detailed case studies and quantifiable outcomes, highlighting the practical benefits and advancements that causality offers in retail banking analytics. We will deepen into how we’ve enhanced our recommendation systems and embedded causal inference in our analytics framework at BBVA AI Factory. Attendees will leave the room with a deep understanding of causal AI applications in the financial sector.

• 
  Álvaro Ibraín Rodríguez Expert Data Scientist, BBVA AI Factory

“Decoding eCommerce” – Facing and solving technical challenges implementing Causal Modeling at Bergfreunde.de” – Find out about the technical challenges (and solutions) Europe’s largest Outdoor eCommerce Retailer Bergfreunde.de (known in the UK as www.alpinetrek.co.uk/) had to face implementing a Causal Model to inform business steering. The focus of this talk will be on exposing the statistical and technical challenges faced, and some of the learnings and solutions implemented.

• 
  Alexander Dzionara Data Scientist, Bergfreunde

• 
  Dimitra Liotsiou Senior Research Data Scientist, dunnhumby

Graph theory and causal analysis are under-used tools in Data Science in general, but we rarely talk about their data discovery potential. By using these techniques, we were able to discover a cost-saving opportunity in logistics that traditional analysis and ML was unable to identify. In addition, the visually approachable nature of graph theory made the opportunity explainable to business stakeholders. This combination of data discovery and interpretability by a non-technical audience makes causal analysis a uniquely powerful tool for quickly solving business problems that require the cooperation of inte

• 
  Robert Nicholls Head of Data Science, BCA

Join an interesting discussion exploring how causal approaches could reshape responsible AI practices, influence policy decisions, and drive innovation in the financial sector. Drawing from their extensive experience in AI ethics, strategy, and data-driven decision-making, our panelists will offer unique insights into the challenges and opportunities presented by causal AI. This forward-looking conversation will focus on the strategic importance of causal AI and its implications for businesses navigating the complex landscape of AI adoption and implementation – providing valuable insights for professionals across industries.

• 
  Maria Axente Head of Public Policy & Ethics, PWC
• 
  Ari Cohen Chief Data Officer EMEA, Macquarie Group

Counterfactuals enable exploration of solutions in causal-AI. In addition to counterfactual exploration of a causal-AI network, we introduce search-and-summary to provide a rich discourse between the engineer and the AI system. Working in the area of operational risk of non-productive time (NPT), we begin from semantic cause-effect chains from root-cause-analysis. These are interwoven in a primarily knowledge-driven causal-AI structure. Population of a large subset of the initial conditional probability tables is through physics models.That partially trained structure is then refined on scenarios from historical operations and expert-driven hypothetical situations. In deployment a behaviour tree architecture enables a neurosymbolic combination of human and IoT establish the right risk model for the right context. The results are analyzed to identify the key paths through the causal-AI, recovering specific semantic cause-effect chains corresponding to root causes. These are used in two ways: firstly they can be passed for summarization by LLMs, where we use roundtrip verification and access to underlying human-verified text; secondly, the causal-AI result is used in a skyline multicriterion search across the historical and hypothetical cases. We present the details of the complete system and highlight the role that causal-AI plus semantic web can play in enabling discourse on operational risk.

• 
  Michael Williams Principal AI Research Scientist, SLB

In retail banking, understanding our clients is critical to offering them the best solutions and products. This talk presents an innovative approach based on causality to calculate optimal financial discounts aiming to benefit both the institution and its clientele. Traditional machine learning techniques fall short in this scenario due to the impossibility of making randomized controlled trials and the presence of confounding bias in observational data. We demonstrate with preliminary results on a real use case how causal inference methodologies can overcome these limitations, enabling us to estimate the effect of financial discounts. Our approach allows for interventions and counterfactual analysis, which is crucial for determining personalized optimal discounts. This presentation will detail our end-to-end causal inference pipeline customized for our problem but adaptable to other use cases. It will cover phases like covariate selection, causal discovery, and effect estimation, among others. We will describe our comprehensive approach, including the application and comparison of multiple causal discovery and estimation algorithms. Furthermore, we will contrast the outcomes of our causal inference pipeline with traditional ML-based methodologies, illustrating how causal inference effectively corrects biases that ML cannot address.

• 
  Clara Higuera Cabañes Lead Data Scientist, BBVA
• 
  Javier Moral Hernández Senior Data Scientist, BBVA