A indústria das telecomunicações é caracterizada pela rápida mudança de ofertas, existindo uma fácil mudança entre operadoras por parte dos clientes. Por conseguinte, é um mercado de forte concorrência onde existe uma forte guerra de preços. Assim, a retenção de clientes é muito importante para as empresas do mercado das telecomunicações, de modo a que estas mantenham os seus clientes e não os percam para as empresas concorrentes. O presente relatório apresenta a descrição das atividades desenvolvidas durante o estágio realizado na direção segmento de consumo residencial, na área de planeamento e projetos estratégicos da PT Comunicações, S.A., o qual durou 4 meses, de 21/04/2014 até 28/08/2014. As atividades desenvolvidas foram a segmentação dos clientes do mercado residencial com base no seu valor para a empresa e no seu comportamento migracional. Os principais objetivos foram pela criação e disponibilização de uma ferramenta de identificação e monitorização de todos os movimentos de clientes, de acordo com a sua subscrição de produtos e serviços. Para alcançar estes objetivos foram analisados dados de faturação de clientes, através da ferramenta SQL Server. De acordo com os objetivo traçados, o estágio permitiu ao estagiário desenvolver as suas competências nas temáticas do SQL, segmentação de clientes e CRM. Numa perspetiva de resultados, procurou-se identificar os principais movimentos de clientes, classificando-os e com isso poder tirar conclusões sobre relações entre a oferta e a procura. Estes resultados permitirão fazer ajustes a nível das ofertas da empresa de modo a responder à procura.

Customer segmentation is vital for companies to cluster their customer base into unique groups that allow for a deeper understanding of such a base. This can provide the companies with ample opportunity to grow revenue and increase customer satisfaction. Nokia’s Advanced Consulting Services (ACS) Team is utilising Factor Analysis to build unique ‘Digital Personas’ for Internet Service Providers (ISPs) with the aims to increase revenue, reduce churn and improve customer experience among other KPIs. This paper highlights the methodology of this clustering, results from a feasibility study and shows how Factor Analysis can be used to segment customer bases and provide recommendations and business intelligence for companies in the telecommunications industry. The methods below produced ten unique digital personas based on data collected from Saudi Telecom (STC). Recommendations such as Fibre to The Home (FTTH) were presented to the client as well as dashboards for their consumption.

This internship report explores the application of the RFM (Recency, Frequency and Monetary) model for customer segmentation in the airport context, focusing on E-Commerce Portuguese customers of ANA – Aeroportos de Portugal. In order to achieve this objective, four interviews were conducted with individuals from the company to gather relevant insights and other perspectives. Before applying the RFM model, comprehensive data preparation was carried out to ensure the accuracy and reliability of the results. By implementing and adjusting the RFM model to the airport context, five distinct segments were identified. The results provided insights into the five segments and enabled the development of targeted marketing actions based on the characteristics of each segment. These findings contribute to understand customer behavior and have practical implications for marketing strategies in the airport industry. The outcomes serve as a foundation for future studies, enhancing understanding of customer behavior and creating target experiences for different passenger segments.

O aumento do tamanho e da dimensionalidade dos repositórios de dados, obriga a que as empresas se adaptem às novas tecnologias e criem novas formas de análise de dados. O clustering é uma das formas mais importantes e provavelmente mais popular de análise de dados. Ao longo deste projeto o clustering será explorado recorrendo ao Self-Organizing Map. Este trabalho irá debruçar-se essencialmente na criação dos clusters dos clientes de uma empresa de logística e distribuição alimentar. O segundo objetivo deste trabalho projeto prende-se a com as conclusões que se podem tirar a partir dos clusters obtidos, isto é, baseado nos diferentes perfis de clientes criados, pretende-se analisar se, é possível melhorar a gestão de stocks, melhorar a estratégia de marketing, antever alguns imprevistos e melhorar a cadeia de supply chain onde a empresa se encontra inserida. Neste âmbito, foi possível concluir que o SOM, é uma ferramenta adequada, para a sua implementação na empresa, uma vez que foi possível diminuir a dimensionalidade e gerar os diferentes grupos de clientes. Com base nestes resultados, concluiu-se que seria possível melhorar a gestão de stocks da empresa e melhorar a estratégia de marketing. Relativamente a toda a cadeia de supply chain onde a empresa se encontra inserida, também esta, será influenciada, uma vez que ao melhorar a gestão de stocks, e aumentando as vendas recorrendo a estratégias de marketing mais eficientes, todos os elementos da cadeia de supply chain sairão beneficiados.

Customer segmentation allows to divide a company’s customers into multiple market segments, enabling the development of customized marketing actions based on each segment’s characteristics. This work describes the application of a customer segmentation approach to the patients of a Portuguese dental company. The approach taken to select the feature subset for the final model was mostly based on the LRFM (length, recency, frequency, and monetary) model, and the monetary variable was split into multiple variables according to the treatment category where the amount was spent. K-Means and Self-organizing maps were used to cluster the company’s patients using these variables, and the results returned by both algorithms are compared. The final solution was obtained with K-Means, and 7 clusters of patients were identified. An overview of the 7 clusters is provided, and possible marketing actions are suggested based on their main characteristics. The results allowed the company to understand how its turnover was distributed across segments, and to develop an initiative to contact the patients belonging to a segment where most of them did not have an appointment in one of the company’s clinics for a long time.

To ensure competitiveness and relevancy in today’s highly digitised world, companies need to ensure that their focus is continuously on the client and on the experience they provide – while not having a negative effect on the organisation’s bottom line. A crucial step to achieving this is to get to know one’s customer base. With the vast amount of data available in a health insurance company, they are able to leverage on unsupervised machine learning techniques to segment their customers. This enables organisations to have a more tailored approach to their customers, identify market growth opportunities and gain competitive advantage.

Although topics of segmentation and loyalty programs are of central importance in the hospitality industry, research about alternative ways to segment loyalty program members is limited and managers tend to rely on traditional segmentation techniques. This study aims to provide a new customer segmentation solution for hotels’ loyalty programs using a data mining approach for identifying and classifying the customers into segments through clustering processes. Guests profiles were assessed with data about 498.655 loyalty members’ of Pestana Hotel Group. The K-means algorithm was applied in order to group the similar guests based on the monetary value of clients and consumption behavior. The goal is to compare the data-driven segments which are based on the customer’s monetary value, brand preferences and demographic data with the loyalty program tiers. The results demonstrate that the widely used tier-based loyalty programs are not optimal and are hiding important features that could be used to improve clients’ segmentation. Findings suggest that some high tier members generate comparatively less revenue for the hotel than lower tier ones. Hence, more efforts should be focused on truly valuable clients. Loyalty programs are not equally suitable for all guests neither for all brands within a hotel group, therefore additional levels of segmentation would be appropriate to match the distinct guests’ behavior. Data mining technologies can be extremely useful in order to support hotel managers in designing a more efficient and valuable loyalty program with tailored strategies and rewards.

Predictive analysis is expected to provide valuable insights, allowing the television channel to direct its sales strategies more assertively, in line with business objectives and maximizing market opportunities. Using machine learning (ML) models to predict cross-selling sales represents an opportunity for many companies to diversify their activities and expand their markets. The data selected was an historical jewelry sales data, demographic characteristics of buyers, and other relevant variables. The mentioned variables aim to identify buying patterns and determine which products are most likely to be purchased together. To maximize the sales of a television channel, this study seeks to help a Brazilian TV jewelry shopping channel to understand the consumption patterns of customers who have become inactive since the last purchase, by analyzing their buying behavior in the previous year. Based on this analysis, it is possible to offer additional products that are relevant and complementary to items already purchased or of interest to customers, generating business opportunities. To achieve this objective, the methodological approach involves applying machine learning algorithms such as clustering, RFM analysis, and a recommendation system to create a predictive model capable of identifying cross-selling or upselling opportunities and considering purchasing patterns and other factors. The analysis uncovered that the largest number of customers is in the Top Customers cluster (30.5%), these customers spend more and buy more frequently, generating high revenue. The cluster in the sequence is the recent customers (29,4%), which have high recency, moderate frequency and monetary scores, so these customers do buy often and have made a purchase recently either. RFM analysis, combined with the recommendation system, provided a powerful way of detecting consumption patterns and recommending relevant products to customers. These techniques helped the company increase customer retention, improve retention, and maximize CLV. This research showed the power of data analysis and machine learning, which could bring in some very key changes in companies' ability to communicate with customers and bring them through more personalized and relevant experiences. Fully accurate identification of consumption patterns and recommendation of related products not only increases customer retention but also helps the company grow faster and be more competitive in the market.

The competitive landscape of the financial sector requires innovative approaches in marketing, prompting banking institutions to integrate data science methodologies. This study aims to bridge the gap in delivering personalized customer experiences by transitioning from traditional rule-based segmentation to data-driven segmentation using machine learning. Focusing on credit card users from a confidential European financial institution, we propose the development of multiple machine learning models to segment customers based on their purchasing behavior. Several clustering algorithms were evaluated, including K-means, Hierarchical Clustering, Gaussian Mixture (GMM), and DBSCAN. The goal is to provide actionable insights to the marketing team, enabling them to make informed decisions regarding marketing campaigns and fostering more personalized customer relationships. We define and utilize various data sources, establish exclusion rules, and apply statistical techniques to ensure data quality and relevance. By employing diverse clustering algorithms, we aim to create dynamic customer segments that can inform personalized marketing strategies. Based on existing literature, it is anticipated that data-driven segmentation will enhance customer satisfaction, retention, and marketing efficiency. This research contributes to the financial marketing field by providing a robust framework for customer segmentation and offering insights into customer behaviors and preferences, which is expected to drive business growth.

This study presents a framework for segmenting Food Delivery Application (FDA) customers based on psychographic and behavioral variables as an alternative to existing segmentation. Customer segments are proposed by applying clustering methods to primary data from an electronic survey. Psychographic and behavioral constructs are formulated as hypotheses based on existing literature, and then evaluated as segmentation variables regarding their discriminatory power for customer segmentation. Detected relevant variables are used in the application of clustering techniques to find adequate boundaries within customer groupings for segmentation purposes. Characterization of customer segments is performed and enriched with implications of findings in FDA marketing strategies. This paper contributes to theory by providing new findings on segmentation that are relevant for an online context. In addition, it contributes to practice by detailing implications of customer segments in an online sales strategy, allowing marketing managers and FDA businesses to capitalize knowledge in their conversion funnel designs.