![](https://expert-mik.ru/wp-content/themes/24author/assets/img/work/work-img.png)
Продвинутая аналитика для прогноза потребительских предпочтений на примере кейса L’Oréal
Целью исследования является анализ поведения потребителей продукции бренда Kiehl’s, принадлежащего компании L’Oréal, для персонализации клиентского опыта за счет выявления схожих паттернов в поведении, построения рекомендательной системы и алгоритма предсказания совершения заказа в будущем.
Задачами проекта являются:
1. Изучить проблему анализа потребительского поведения и проанализировать актуальные теоретические и практические подходы к проблеме;
2. Осуществить обзор литературы об использовании продвинутых аналитических инструментов для персонализации клиентского опыта на примере кластеризации и построения рекомендательной системы;
3. Определить методы и способы осуществления кластеризации потребителей на основе схожести их поведения, построения рекомендательной системы и прогнозного алгоритма;
4. Провести разведочный анализ данных, осуществить кластеризацию с помощью выбранного метода, создать рекомендательную систему и прогнозный алгоритм;
5. Оценить экономическую эффективность внедрения полученных решений компанией, описать первые этапы внедрения систем.
В результате работы была проведена кластеризация клиентов компании, построена рекомендательная система для товаров бренда, а также был разработан алгоритм, прогнозирующий совершение покупки клиентов в краткосрочной перспективе. Полученные результаты были оценены с точки зрения экономической эффективности, разработанные решения были переданы компании.
INTRODUCTION 8
CHAPTER 1. INTRODUCTION TO CONSUMER BEHAVIOR ANALYSIS 10
1.1. Process and motives of consumer behavior analysis 10
1.2. Role of machine learning in consumer behavior analysis 14
1.3. Advanced techniques to analyse a consumer on the example of business cases 18
1.4. Typical challenges of consumer behavior analysis and recommendation system creation 27
1.5. Conclusion 31
CHAPTER 2. CONSUMER BEHAVIOR ANALYSIS METHODS REVIEW 32
2.1. Methods of data clustering
2.2. Methods of recommendation system development 2.3. Chosen methods for the project
2.4. Conclusion
CHAPTER 3. DATA ANALYSIS AND APPLICATION OF CHOSEN
3.1. Exploratory data analysis 3.2. Consumers clustering
3.2.1. Findings on consumer clustering 3.3. Recommender system creation
3.3.1. Utilization of the data set
3.3.2. Model description
3.3.3. Findings on recommendation system 3.3.4. Potential gaps of the recommender
3.4. Predictive model on customer’s potential purchase 3.4.1. Data preprocessing
3.4.2. Findings on the purchase prediction
3.5. Recommendations for further business implication 3.6. Future implementation process
3.7. Conclusion
METHODS 52
CONCLUSION 94 LIST OF REFERENCES 101 APPENDICES 106
This master’s thesis is a research project for one of the largest international FMCG players in Russia and in the worl – L’Oréal company, and particularly for its brand Kiehl’s. The company operates in the sector of beauty care products (including decorative cosmetics, perfumery, skin care and hair care) and manages a portfolio of over 30 brands. The company has 4 main divisions: consumer products (decorative and care cosmetics of low and medium price segment, available to a wide audience), luxury products (cosmetics and care products of the high price segment), active cosmetics (medical and professional skin care products) and professional products (various hair products intended for use by professionals in beauty and hairdressing salons). The Kiehl’s brand itself, which this study is performed for, belongs to the luxury products division.
Most previously existing papers on the topic of advanced techniques for consumer behaviour analysis were mainly attempting to focus on short-term trends of consumer behavior, while longer relationships between a company and a client tended to be a blind spot of most researchers. Furthermore, the idea of bringing comprehensive analytics to the cosmetics industry was not developed and embodied to a sufficient extent, since only few companies in this segment were able to implement the findings and align customized solutions with their business objectives. Thus, the above aspects result in a significant gap in investigation of advanced analytics and recommender systems within the FMCG cosmetics niche with building the ready-to-apply model introducing the main academic gap, which gives the team of researchers the space to conduct this research.
The research is motivated by several factors. Firstly, there is a massive upcoming trend for customizing the user experience, therefore, lack of product or service customization is a significant competitive disadvantage for business. Business case analysis demonstrates that customized approach is an effective way to build closer connections with customers and foster their loyalty. Moreover, availability of big volumes of customer data to the companies and lowering costs of storage and processing make the analysis of consumer behavior more technically accessible to various companies. Finally, predictive analytics becomes more and more in demand in business, since it leads to optimized activities of production, marketing, logistics and R&D units, thus affecting a company’s overall financial performance.
8
This study is an attempt to close several revealed gaps. From the academic perspective, there is a limited number of studies looking at the application of machine learning methods to the analysis of consumer behavior. Moreover, there is also a significant lack of research on the application of theoretical approaches to the creation of recommender systems for the Russian FMCG industry. From the perspective of business, it was revealed that currently used advanced analysis techniques and tools are very narrowly applied within short-term solutions within L’Oréal and present a fragmented technology, which cannot be fully deployed and lead to observable effects. In addition, the company does not cover long-term behavior prediction of FMCG or cosmetics client’s future purchases. Finally, no one designed and implemented any customer behaviour analysis system within the particular brand of Kiehl’s.
The goal of the project is to analyze consumer behavior in order to personalize the customer experience by using recommendation models, clustering of clients or an algorithm for predicting the order in the future. To address the issue of building recommendations for particular brand of L’Oréal – Kiehl’s – as well as to highlight key parameters that influence the customers’ choices, demonstrating the most effective way to make it a lasting solution for increase in the brand’s performance, the following research questions (RQ) were formulated:
RQ 1: What are the main features and areas of application of consumer behavior analysis in today’s business environment?
RQ 2: What methods are used to solve the problems of consumer behavior analysis and what methods can be chosen for Kiehl’s case?
RQ 3: What similarities can be found in Kiehls’ consumer behavior, and how can we use this to make the customer experience more personalized?
RQ 4: Is it possible to recommend to Kiehl’s customer an item from the proposed range of products that is likely to be bought?
RQ 5: How, based on the previous purchase history, identify if Kiehl’s client is about to make an order within the next period (one month)?
In the first chapter of the study, the paper would consider consider the motivation, specificities and procedure of conducting consumer behavior analysis, as well as highlight the issue of using machine learning algorithms for consumer analysis, consider in detail the business
9
cases of large companies involved in consumer behavior analysis, and describe possible difficulties. The second chapter is devoted to a theoretical review of existing methods used to carry out the analysis, that would also assist in a decision on tools used for this investigation later when it comes to the empirics.. The third chapter describes the practical application of the methods selected in the second chapter on data provided by the brand’s representatives specifically for this research. Then a closer look at details in data provided would reveal its peculiarities, and after an initial analysis of the data is conducted, the research moves to consumers clusterization, creation of a recommendation system and also an algorithm that predicts whether a user is about to make an order for the brands’ products within next month.
Thus, the expected results are answers to the research questions outlined above, highlighted consumer clusters, a reliably working recommendation system, and a purchase predictor. There would also be a set of managerial recommendations provided for the company.
![](https://expert-mik.ru/wp-content/themes/24author/assets/img/signup-document.png)
Последние выполненные заказы
Хочешь уникальную работу?
Больше 3 000 экспертов уже готовы начать работу над твоим проектом!