ilayzvulun/gym_churn

README Project Overview This project performs an Exploratory Data Analysis (EDA) on a gym membership dataset in order to understand patterns that may influence customer churn. The analysis focuses on identifying whether weekly training frequency has a meaningful effect on churn and how different variables in the dataset may support or complement this understanding. The dataset includes demographic attributes, behavioral training patterns, contractual details, and churn indicators for thousands of gym members.

Dataset Features The dataset contains several types of attributes: Demographic Features • Age • Gender • Height • Weight • BMI Behavioral and Engagement Features • Avg classes per week • Days since last visit • Additional monthly charges • Membership type Contract Related Features • Remaining months on contract • Current membership type Target Variable • Left (0 for active members, 1 for churned members)

Research Question How does a customer’s weekly training frequency affect the likelihood of churn? This question guides the entire analysis and frames the purpose of each visualization and interpretation.

Data Cleaning and Preparation Before performing the EDA, the dataset underwent several preprocessing steps: • Verified that no values were missing • Ensured all categorical fields contained valid entries • Checked for outliers and confirmed they were realistic • Verified that all numeric features were within expected ranges • Confirmed dataset consistency across all variables The dataset was then used to generate visual insights related to training habits and churn behavior.

Graphs, Questions, Answers, and Insights

1. Distribution of Age

Question How is the age of gym members distributed, and can age groups help identify potential differences in training habits that may relate to churn? Answer and Insight The age distribution shows that most members are between 24 and 32 years old. While age does not directly measure training frequency, it can help identify groups that may behave differently in their training habits. Younger members often display less consistent exercise routines, which may increase their likelihood of lower weekly frequency and therefore raise their risk of churn. This makes the age distribution useful for understanding which segments may be more prone to reduced training and potential churn.

2. Distribution of Weekly Class Frequency

Question What is the distribution of weekly training frequency among members, and what does it reveal about engagement levels related to the research question? Answer and Insight The chart shows that most members train between 1 and 3 times per week, with a clear peak around 2 sessions. This distribution provides an essential baseline for the research question because it reveals that many members already operate at relatively low frequency. Since the research question examines how training frequency influences churn, understanding this baseline helps identify which ranges of training activity might be associated with higher churn risk.

3. Training Frequency by Churn Status

Question Does weekly training frequency differ between members who stayed and members who churned, and how does this difference support the research question? Answer and Insight The boxplot shows a clear difference between active and churned members. Customers who churned tend to train less often, typically between 0 and 2 times per week, while active members show higher and more consistent training frequency. This directly supports the research question by demonstrating that lower weekly exercise frequency is strongly associated with a higher likelihood of churn.

4. Frequency vs Additional Charges

Question Is there a relationship between weekly training frequency and additional monthly charges, and can spending behavior help explain churn related patterns? Answer and Insight The scatterplot shows no strong relationship between training frequency and additional monthly charges. This means that spending patterns do not appear to change based on how often a member trains. This supports the research question by showing that training frequency is an independent behavioral factor. Since additional spending does not influence or explain frequency, any effect on churn is more likely connected directly to how often the customer trains.

5. Distribution of Remaining Contract Months

Question How many months do members have left in their contracts, and could different contract stages influence training consistency and churn? Answer and Insight The distribution of remaining contract months shows that many members have either very short or very long time left in their membership. While this variable does not directly reflect training frequency, it can influence member behavior. For example, customers nearing the end of their contract who already train infrequently may be more likely to churn. Therefore, this graph supports the research question by highlighting groups whose contract stage may be associated with lower training frequency and higher churn risk. Overall Insights • Training frequency is one of the most meaningful behavioral indicators of churn. • Members with low weekly activity levels are significantly more likely to leave. • Other variables, such as additional charges or remaining contract months, do not directly cause churn but may provide useful context. • Understanding frequency patterns across age groups and contract stages can help identify high risk segments.

Decisions and Recommendations Based on the analysis, the following actions may help reduce churn: • Encourage members with low frequency to increase weekly training habit • Provide personalized follow up for members training 0 to 1 times per week • Improve engagement using reminders, onboarding support, and class variety • Monitor members nearing the end of their contract who already show low activity • Treat training frequency as a core KPI when evaluating member retention strategy

Project Files • Dataset uploaded to HuggingFace • Full notebook including all code and visualizations • This README summarizing the analysis and outcomes

https://youtu.be/y_lH8Mg0XhM