The cosmetics and skincare industry is characterized by intense competition, fragmented supply, and rapidly evolving consumer preferences. In this context, the ability to understand and anticipate repurchase behavior represents a fundamental strategic advantage for firms operating in the sector. This thesis develops an integrated analysis of customer retention and churn for a skincare company, using transactional data from the company’s ecommerce platform over the period 2021 to 2025. The objective is twofold: on the one hand, to identify the factors influencing the probability of repurchase and to build accurate predictive models; on the other hand, to translate these predictions into operational marketing strategies aimed at profit maximization. The analysis is structured around three complementary methodological approaches. First, survival analysis using Cox models, both standard and LASSO penalized, makes it possible to model time to repurchase while appropriately accounting for data censoring and identifying the most relevant covariates. Second, several predictive models are implemented for the binary classification of repurchase behavior, including LASSO penalized logistic regression, random forests, gradient boosting, support vector machines, and deep neural networks. Finally, the Profit and AUC focused Prescriptive Analytics Method, PAM, is applied to translate predicted probabilities into operational decisions that maximize the expected profit of retention campaigns, taking into account heterogeneity in customer lifetime value and intervention costs. The findings highlight how integrating multiple analytical techniques makes it possible to move beyond a purely predictive dimension, transforming forecasts into data driven operational decisions aimed at maximizing the return on marketing investments.
Il settore della cosmetica e della skincare è caratterizzato da elevata competitività, frammentazione dell'offerta e rapida evoluzione delle preferenze dei consumatori. In questo contesto, la capacità di comprendere e anticipare il comportamento di riacquisto rappresenta un vantaggio strategico fondamentale per le aziende del settore. Questo lavoro di tesi sviluppa un'analisi integrata della retention e del churn nella clientela di un'azienda di skincare, utilizzando dati transazionali provenienti dalla piattaforma ecommerce nel periodo 2021-2025. L'obiettivo è duplice e consiste, da un lato, nell'identificare i fattori che influenzano la probabilità di riacquisto e nel costruire modelli predittivi accurati e, dall'altro, nel tradurre tali previsioni in strategie operative di marketing orientate alla massimizzazione del profitto. L'analisi si articola su tre approcci metodologici complementari. Innanzitutto, l'analisi di sopravvivenza mediante modelli di Cox, sia standard che penalizzati con lasso, permette di modellare il tempo al riacquisto considerando appropriatamente la censura dei dati e identificando le covariate più rilevanti. Successivamente, vengono implementati diversi modelli predittivi (regressione logistica con penalizzazione lasso, foreste casuali, gradient boosting, support vector machine, reti neurali profonde) per la classificazione binaria del comportamento di riacquisto. Infine, viene applicato il PAM (Profit- and AUC-focused prescriptive analytics method) per tradurre le probabilità predette in decisioni operative che massimizzano il profitto atteso delle campagne di retention, considerando l'eterogeneità del customer lifetime value e i costi di intervento. Il lavoro evidenzia come l'integrazione di diverse tecniche di analisi consenta di superare la sola dimensione predittiva, traducendo così le previsioni in decisioni operative data-driven orientate alla massimizzazione del ritorno sugli investimenti di marketing.
Previsione del riacquisto nel settore cosmetico: dal modello predittivo alla strategia di marketing
BISSO, FEDERICA
2025/2026
Abstract
The cosmetics and skincare industry is characterized by intense competition, fragmented supply, and rapidly evolving consumer preferences. In this context, the ability to understand and anticipate repurchase behavior represents a fundamental strategic advantage for firms operating in the sector. This thesis develops an integrated analysis of customer retention and churn for a skincare company, using transactional data from the company’s ecommerce platform over the period 2021 to 2025. The objective is twofold: on the one hand, to identify the factors influencing the probability of repurchase and to build accurate predictive models; on the other hand, to translate these predictions into operational marketing strategies aimed at profit maximization. The analysis is structured around three complementary methodological approaches. First, survival analysis using Cox models, both standard and LASSO penalized, makes it possible to model time to repurchase while appropriately accounting for data censoring and identifying the most relevant covariates. Second, several predictive models are implemented for the binary classification of repurchase behavior, including LASSO penalized logistic regression, random forests, gradient boosting, support vector machines, and deep neural networks. Finally, the Profit and AUC focused Prescriptive Analytics Method, PAM, is applied to translate predicted probabilities into operational decisions that maximize the expected profit of retention campaigns, taking into account heterogeneity in customer lifetime value and intervention costs. The findings highlight how integrating multiple analytical techniques makes it possible to move beyond a purely predictive dimension, transforming forecasts into data driven operational decisions aimed at maximizing the return on marketing investments.| File | Dimensione | Formato | |
|---|---|---|---|
|
Bisso_Federica.pdf
Accesso riservato
Dimensione
8.31 MB
Formato
Adobe PDF
|
8.31 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/105766