Abstracts

NOVARTIS   Abstract n°1 Complaint trend evaluation in a pharma industry  

The objective of this project is to develop a robust statistical model aimed at analyzing trends and predicting complaints within a packaging site. The project will involve the collection and analysis of historical complaint data to identify patterns and key factors contributing to customer dissatisfaction. By leveraging advanced statistical techniques and machine learning algorithms, the model will be designed to forecast future complaint trends, enabling proactive measures to mitigate potential issues.

The scope of the project includes:

1. Data Collection and Preprocessing: Gathering historical complaint data from various sources within the packaging site, followed by data cleaning and preprocessing to ensure accuracy and consistency; the process is facilitated by the site team.

2. Exploratory Data Analysis (EDA): Conducting a thorough analysis to uncover underlying patterns, correlations, and anomalies in the complaint data.

3. Model Development: Utilizing statistical methods and machine learning algorithms to build a predictive model. Techniques such as time series analysis, regression analysis, and classification algorithms will be explored.

4. Model Validation and Testing: Evaluating the model's performance using historical data and refining it to improve accuracy and reliability.

5. Implementation and Monitoring: Deploying the model within the packaging site’s operational framework and establishing a monitoring system to track its performance and update it as needed.

The anticipated outcome of this project is a predictive tool that will enable the packaging site to anticipate and address potential complaints, thereby enhancing customer satisfaction and operational efficiency.

NOVARTIS Abstract n°2 Packaging yield evaluation.

This project aims to evaluate the packaging yield in a packaging site by analyzing various factors that influence the yield of packaging batches. The primary objective is to identify and quantify the impact of different variables such as packaging technology, product brand, packaging configuration, number of packs per run, and production sequences (product mix) on the overall packaging yield.

The scope of the project includes:

• Data Collection and Preprocessing: Gathering detailed data on packaging batches, including information on packaging technology, product brand, packaging configuration, number of packs per run, and production sequences. This data will be cleaned and preprocessed to ensure accuracy and consistency.

• Exploratory Data Analysis (EDA): Conducting a comprehensive analysis to identify patterns, correlations, and potential outliers in the data. This step will help in understanding the relationships between different variables and their impact on packaging yield.

• Model Development: Developing statistical models to evaluate the yield of packaging batches. Various techniques such as regression analysis, ANOVA, and machine learning algorithms will be employed to build robust models that can accurately predict packaging yield based on the identified variables.

• Model Validation and Testing: Validating the developed models using historical data and testing their performance to ensure reliability and accuracy. This step will involve refining the models to improve their predictive capabilities.

• Implementation and Monitoring: Implementing the models within the packaging site's operational framework and establishing a monitoring system to track their performance and update them as needed.

The anticipated outcome of this project is a comprehensive understanding of the factors affecting packaging yield and the development of predictive models that can help optimize packaging processes. This will lead to improved efficiency, reduced waste, and enhanced overall productivity in the packaging site.

TOOLSGROUP Modeling the demand in retail within a complex pricing landscape

Demand forecasting plays a crucial role in supporting decisions for policy makers across different industries [1]. Predicting the demand for a product in retail is generally a highly challenging task for numerous reasons, many of which are sector-specific, such as the frequent introduction of new products, a short life cycle, returns and a strong seasonality in fashion business [2] or perishable products in food industry [3].

Among the various factors influencing the demand, pricing is one of the most important, as it is ubiquitous across retail sectors and is a factor over which the retailer has direct control. The notion of price elasticity has been historically used to quantify the influence of price on the demand for a product [4]. In practice, elasticity should be understood more broadly, capturing not only the effect of an item’s own price on its demand, but also the influence of other products’ prices (cross-elasticity) [5], and of the recent history of pricing and promotions [6].

We propose developing a demand model that captures the effect of a complex pricing landscape, modeling the effect of price changes on the demand of the product which changes price (self-elasticity), but also focusing on cross-elasticity and recurrent promotions. The successful model should provide quantitative instruments to deal with crucial and impactful questions for revenue optimization, such as “What is the effect of changing the price of a product on its demand and on the demand of other available products?” and “What is the effect of promoting a product on its demand during the promotion and outside the promotion period?”. We will provide sample data to tackle the problem, and the goal is to code a working prototype of the model. However, we will also value the theoretical approach towards answering these crucial questions.

[1] Ingle, Chaitanya, et al. "Demand forecasting: Literature review on various methodologies." 2021 12th International Conference on Computing Communication and Networking Technologies (ICCCNT). IEEE, 2021.

[2] Swaminathan, Kritika, and Rakesh Venkitasubramony. "Demand forecasting for fashion products: A systematic review." International Journal of Forecasting 40.1 (2024): 247-267.

[3] Van Donselaar, Karel H., et al. "Analysis and forecasting of demand during promotions for perishable items." International Journal of Production Economics 172 (2016): 65-75. 

[4] Özer, Özalp, and Robert Phillips, eds. The Oxford handbook of pricing management. Oxford University Press (UK), 2012.

[5] Sarkar, Biswajit, et al. "Is the system reliability profitable for retailing and consumer service of a dynamical system under cross-price elasticity of demand?." Journal of Retailing and Consumer Services 75 (2023): 103439.

[6] Abolghasemi, Mahdi, et al. "Demand forecasting in the presence of systematic events: Cases in capturing sales promotions." International Journal of Production Economics 230 (2020): 107892.

FONDACO GROUP  Probabilistic assessment of macro-financial scenarios 

This research project aims to develop a quantitative framework for the probabilistic assignment of future macro-financial scenarios, which are ex ante defined by the portfolio manager.

At a given time t, the manager provides a discrete set of scenarios representing potential evolutions at time t+1 of a range of macroeconomic and financial variables such as equity indices, interest rates, credit spreads, and other relevant indicators. Each scenario is characterized by a set of expected values for all considered variables, constructed according to distinct macroeconomic narratives (e.g., expansion, slowdown, exogenous shock).

The objective is to construct a model that can assign a probability of occurrence to each scenario in a manner consistent with:

• the historical time series of macro-financial variables observed up to time t;

• the underlying dependency structure and co-movement among these variables.

Since the proposed scenarios significantly influence the expected returns of portfolio assets and, by extension, investment decisions. It is essential to rely on a robust quantitative tool. Such a model must translate the portfolio manager’s qualitative views into a coherent and internally consistent probabilistic system, thereby enhancing the rigor and transparency of the investment decision-making process.