Applying machine learning to user-generated content to identify Nespresso customers’ pain points
SONAR|HES-SO
- Torche, Nicolas
- Meidell, Jan Erik (Degree supervisor)
- Genève : Haute école de gestion de Genève
72 p.
Bachelor of Sciences HES in International Business Management: Haute école de gestion de Genève, 2023
English
This thesis is a practical application of machine learning, more specifically natural language processing (NLP), to user-generated content related to the company Nespresso. It covers the entire process necessary to analyse user-generated content, including data collection, data pre-processing, data labelling, algorithm training and selection, and analysis. The objective of this thesis is to identify the most frequently expressed pain points of Nespresso customers.
The data was collected from the social media platform Twitter, where the most recent 400’000 tweets regarding Nespresso were collected. The data was then pre-processed, and a portion of it was pre-labelled with a large language model. It was then manually reviewed in order to create a training dataset for the machine learning algorithms. The algorithms were trained with simple under-sampling, data augmentation with under-sampling, and data augmentation with over-sampling. After selecting the best performing model, pain point tweets were identified across the entire tweet dataset. These pain point tweets were then analysed with the help of three NLP tools, namely N-grams, LDA and BERTopic. After analysing the results with the help of the three NLP tools, the following most frequently mentioned pain point categories were identified:
• Ordering & Delivery
• Machine & Capsules
• Customer Service
While other pain point categories were identified, these were the most frequently mentioned ones. For each category, multiple anonymised tweet examples are shown, to provide context and evidence for the findings.
The ordering and delivery category includes pain points that are typically related to late deliveries, the delivery person not collecting capsule recycling bags, customers not receiving an order tracking number, and having their package lost and never reaching their address.
The most frequent pain points related to machines are technical issues such as the machine breaking or stopping to work, leaking, loudness and descaling. Another interesting pain point is that customers tend to forget to place a cup under the machine when preparing their coffee.
The data was collected from the social media platform Twitter, where the most recent 400’000 tweets regarding Nespresso were collected. The data was then pre-processed, and a portion of it was pre-labelled with a large language model. It was then manually reviewed in order to create a training dataset for the machine learning algorithms. The algorithms were trained with simple under-sampling, data augmentation with under-sampling, and data augmentation with over-sampling. After selecting the best performing model, pain point tweets were identified across the entire tweet dataset. These pain point tweets were then analysed with the help of three NLP tools, namely N-grams, LDA and BERTopic. After analysing the results with the help of the three NLP tools, the following most frequently mentioned pain point categories were identified:
• Ordering & Delivery
• Machine & Capsules
• Customer Service
While other pain point categories were identified, these were the most frequently mentioned ones. For each category, multiple anonymised tweet examples are shown, to provide context and evidence for the findings.
The ordering and delivery category includes pain points that are typically related to late deliveries, the delivery person not collecting capsule recycling bags, customers not receiving an order tracking number, and having their package lost and never reaching their address.
The most frequent pain points related to machines are technical issues such as the machine breaking or stopping to work, leaking, loudness and descaling. Another interesting pain point is that customers tend to forget to place a cup under the machine when preparing their coffee.
- Language
-
- English
- Classification
- Economics
- Notes
-
- Haute école de gestion Genève
- International Business Management
- hesso:hegge
- Persistent URL
- https://folia.unifr.ch/global/documents/327704
Statistics
Document views: 93
File downloads:
- IBMBT_24_TORCHE_Nicolas_VF.pdf: 221