Departamento de Ingeniería Industrial de la Universidad de Chile Facultad de Ciencias Físicas y Matemáticas de la Universidad de Chile

Working Papers y Papers

Gestión de Cadenas.

We develop a structural demand model that endogenously captures the effect of out-of-stocks on customer choice by simulating a time-varying set of available alternatives. Our estimation method uses store-level data on sales and partial information on product availability. Our model allows for flexible substitution patterns which are based on utility maximization principles and can accommodate categorical and continuous product characteristics. The methodology can be applied to data from multiple markets and in categories with a relatively large number of alternatives, slow moving products and frequent out-of-stocks (unlike many existing approaches). In addition, we illustrate how the model can be used to assist the decisions of a store manager in two ways. First, we show how to quantify the lost sales induced by out-of-stock products. Second, we provide insights on the financial consequences of out-of-stocks and suggest price promotion policies that can be used to help mitigate their negative economic impact, which run counter to simple commonly used heuristics.

Keywords: aggregate demand estimation; Bayesian methods; choice models; data augmentation; inventory management; out-of-stocks; retailing.

Descargar PDF

En este trabajo se propone un modelo de estimación de demanda basado en los atributos que describen una categoría de productos (por ejemplo marcas, tamaños, sabores, etc.). El modelo se basa en crear relaciones analíticas entre los productos y los atributos que los componen, de esta manera basta estimar los parámetros del modelo solo utilizando información de atributos y a partir de cálculos se llega a los parámetros a nivel de SKU.

El modelo se valida con un conjunto de datos simulados para luego ser calibrado con 181 días de datos reales de dos subcategorías distintas: pastas de diente y lavalozas. Entre los resultados se encuentra que los ajustes a nivel de atributos son considerablemente superiores a los obtenidos a nivel de SKU. Por otra parte, los parámetros que entrega el modelo se utilizan para recobrar elasticidades a nivel de atributos y a nivel de SKU, y además se utilizan para realizar estimaciones de demanda, las que tienen un error promedio porcentual del orden de un 30%. Dichos pronósticos superan a los obtenidos al utilizar otras técnicas, como alisamiento exponencial simple, y alisamiento exponencial con estacionalidad, entre otros. Descargar PDF

Gestión de Sala.

La principal causante de faltantes de mercadería en góndolas (FMG) en supermercados es la deficiente calidad de la información almacenada en los sistemas de inventario.

En el presente trabajo se propone un modelo de clasificación que determine si un producto se encuentra en estado de FMG. El modelo no ocupa información de inventario sino básicamente los datos generados por los puntos de venta.

El proyecto consta de un levantamiento de información, pre-procesamiento de datos y pruebas de modelos de redes neuronales Multi-Layer Perceptron. Los resultados indican que es posible detectar eventos de quiebre de stock en góndolas a través de la información transaccional, facilitando de esta forma la detección temprana de productos en estado faltante.

Descargar PDF.

Gestión de Categoría.

El proyecto consiste en desarrollar un software que permita capacitar a los encargados de la administración de categorías en supermercados, a través de la simulación de las ventas semanales para una categoría de productos, basándose en las decisiones de marketing tomadas para la semana en curso y la categoría en estudio. Descargar PDF

Traditional marketing and economic literature including search costs in purchase behavior analyze the case of symmetric equilibrium where all retailers fix the same price. This type of equilibrium outcome does not provide useful insights for the case where some intrinsic competitive advantage of some players allow them to offer lower prices. This article explore consumer search processes describing alternative modeling approaches. Then, we propose an asymmetric equilibrium model in a duopoly to explain how specialty retailer could use a higher assortment flexibility to defend from this type of mass merchandiser competition. Finally we compare equilibrium outcomes in several scenarios and propose some extensions to the model.

Descargar PDF.

Gestión de Clientes.

Customer churn detection is relevant for any company: the cost of acquiring a good new customer is far greater than the cost of retaining a good “old” one. The proposed methodology characterizes the churn process of customers using their transactional data. The dynamics of the situation are modeled with Markov chains, where each state represents a group of clients with similar Recency, Frequency and Monetary value attributes. The
methodology characterizes the “path” that a customer leads to churn, besides estimating the churn probabilities in every period. Based on information analyzed, the methodology proposed detects how customers are moving slowly towards inactivity. This knowledge is valuable to managers because it gives them a timely signal they can use to define preventive actions to retain valuable customers.

Descargar PDF.

Lifetime Value Estimation Based on Transactional Data

The lifetime value

of a customer is the present value of all the future earnings that he will represent for the firm. The estimation of this indicator is central to the customer relationship management concept. It empowers managers with a tool they can use determine how much time, effort and money they can invest on their customers. In retail, Recency, Frequency and Monetary Value (RFM) attributes can be used to characterize a customer’s transactional behavior. Models such as the Pareto/NBD (and other simpler ones) use this information to estimate the customer’s expected future behavior.

Specifically, the number of future purchases, their value, and the activity probability can be estimated to generate a customer’s lifetime value. The proposed lifetime value estimation methodology is applied on a cohort of customers of a wholesale supermarket chain. The models prove to be useful in predicting the behavior of segments of customers, specifically on which groups are expected to change their behavior in the future.

Descargar PDF.