AI for Marketing – how digital technologies are studied and applied

The School of Economics and Management at HSE University pays special attention to international relations, expanding the geography of interaction. Partner universities are interested in joint educational projects and scientific research, and they are also eager to share their experience and knowledge. Thus, the Brazilian university UniCamp (https://unicamp.br/) held a two-hour webinar for students of the educational program "International Bachelors in Business and Economics" on the topic "Artificial Intelligence in Marketing."

UniCamp is a research university located near São Paulo, Brazil. The university is considered one of the best universities in Brazil and Latin America and conducts about 15% of Brazilian research. The university has several research centers. For example, the research hub studying artificial intelligence and computer cognitive science (H.IAAC – Hub for Artificial Intelligence and Cognitive Architectures) conducts research in the areas of Distributed Learning, Learning in Cognitive Architectures, Cognitive Architectures, Research Management, AI for Finance, AI for Marketing, Natural Language Processing, and Knowledge Representation.

The academic director of the program, Natalia Khazieva, invited Professor Marcelo da Silva Reis and his team of researchers to share their views on the current state, pitfalls, and opportunities in marketing during the active development of generative artificial intelligence. During the two-hour webinar, which was attended by 115 program students, colleagues from the Brazilian university shared their research on recommendation systems, customer segmentation approaches, and methods for predicting customer churn. Professor Marcelo da Silva Reis spoke about his team and highlighted the main research directions of the center: analysis of customer behavior based on nontraditional data, interpretability in churn prediction, and the privacy and ethics of the marketing ecosystem in Brazil.

Professor Marcelo da Silva Reis

The 2-hour webinar offered by Fillipe, Mariana and me was a great experience from our point of view! We were able to show to HSE students some of our ongoing, cutting-edge research in AI for Marketing, and also to have some insightful discussion after the presentations. We are thankful to Natalia Khazieva for the opportunity of this webinar, which also might be the beginning of further collaboration initiatives between researchers and students from both HSE and University of Campinas.

In Professor Marcelo da Silva Reis's team, there are several young researchers. For example, Filipe Santos develops recommendation systems and frameworks for customer segmentation. It was surprising to learn that colleagues often use the names of various birds (for example, seriema and arara) as names for systems and frameworks.

Fillipe dos Santos Silva, researcher

Participating in the webinar was a very enriching experience. I presented ARARA, a multi-agent framework we developed for conversational recommendation with LLMs. The idea came from the need to overcome LLM limitations in personalized, multi-turn interactions. The event was very well organized and created an open space for technical discussions and idea exchange. I especially enjoyed the questions and overall engagement from the participants. I believe both our team and the audience left with new perspectives for future research.

Another researcher at the center, Mariana Ferreira, spoke about studying customer churn in various areas as one of the ways to manage customer relationships (CRM). Mariana noted that the main reasons for customer churn are poor onboarding, weak customer relationships, and poor customer service. To address these issues, machine learning and deep learning are often used. Mariana studies customer churn in the banking sector, retail, and telecom. At the same time, there is a particular scientific interest in the possibility of testing the customer churn prediction models developed by Brazilian researchers in various markets, including the Russian market.

Mariana Ferreira, researcher

It was a great experience presenting my research and the advances in IA for churn prediction for the participants. We discussed churn prediction, a theme that gained a lot of attention in the marketing areas of big companies to preserve their clients' satisfaction and by consequence, their profit. I also showed what are the tools and techniques in the areas of Machine Learning and IA are being used to solve the problem and how we can evolve the methodology, applying causal methods to deliver to the business areas more precise and robust estimates.

The competencies and knowledge of the researchers from UniCamp were also interesting to the students of the "International Bachelor in Business and Economics" program, as they complement the courses offered in the program (there are several disciplines dedicated to marketing and the use of digital tools in various fields) and are useful for preparing term papers.

Yulia Kvon, a third-year student of the "International Bachelor’s Program in Business and Economics"

The webinar on the application of AI in marketing turned out to be very useful and interesting, as it addressed truly relevant business topics. Especially memorable were the studies related to causal analysis, as they not only allow for forecasting but also help understand what specifically causes churn and provide the opportunity to take preventive measures in advance. For myself, I noted that high model accuracy is not always important.

In addition, issues of ethics and risks associated with the widespread use of AI were raised, which is especially important in the context of the active integration of AI into everyday life and business. Overall, the webinar provided an understanding of how scientific developments can be applied in marketing today. Such webinars are very valuable as they show how what we study at the university works in real business.

In 2024, a team of researchers from the H.IAAC center published the article "Customer churn prediction in imbalanced datasets with resampling methods: A comparative study" in the journal Expert Systems With Applications 246. The article is dedicated to customer churn prediction (CCP) in the context of imbalanced datasets. The authors examine three public datasets from the telecommunications, online retail, and banking sectors, where there is a significant class imbalance (for example, few customers leave compared to those who stay). Professor Marcelo da Silva Reis and his team will continue research in this area and are interested in conducting joint educational events and research with the School of Economics and Management at HSE University – St. Petersburg in the next academic year.

We thank Pavel Khan for preparing the informational materials.