RideShare Customer Analysis

This project analyzes ride-sharing customer data to calculate various metrics for customer engagement and retention. The project has two primary objectives:

1. Calculate hourly, daily, weekly, and monthly averages of customer ride counts.
2. Compute week-on-week cohorts for analyzing customer retention across weeks.

Dataset

The dataset can be accessed here: Dataset Link

The data contains information on customer rides, including timestamps and customer identifiers.

Objectives

1. Aggregated Averages Calculation

The first part of the project calculates hourly, daily, weekly, and monthly averages of ride counts for each customer. These averages are stored and made accessible for querying purposes. This allows for easy access to ride patterns and engagement metrics on different time scales.

Expected Output

• Hourly Average: Average number of rides per hour for each customer.
• Daily Average: Average number of rides per day for each customer.
• Weekly Average: Average number of rides per week for each customer.
• Monthly Average: Average number of rides per month for each customer.

2. Week-on-Week Cohort Analysis

The second part of the project computes week-on-week cohorts. This analysis helps in understanding customer retention, showing how many customers from each starting week remain active in the following weeks.

Input

• Starting Week: The initial week for cohort analysis.
• Number of Weeks: The number of consecutive weeks to consider in the cohort analysis.

Sample Output

For an input starting with the week of 1/12/2019 and covering 3 weeks:

           1/12/2019   8/12/2019   16/12/2019
1/12/2019     2         2            2
8/12/2019     null      3            3
16/12/2019    null      null         4

Explanation

• 2 customers were active in the week of 1st December. Of these:
• 3 new customers were active in the week of 8th December. Of these:
  • 2 returned from the week of 1st December.
• 4 customers were active in the week of 16th December. Of these:
  • 2 returned from the week of 1st December.
  • 3 returned from the week of 8th December.

Implementation

Technology Stack

• Apache Spark: Used for large-scale data processing, supporting both Python and Scala for flexible development.
• Scala/Python: Core languages for the Spark programs to calculate averages and cohort analysis.

Approach

1. Data Preprocessing:

   • Extract and label each timestamp with relevant time segments (hour, day, week, month) for aggregation.

2. Calculating Averages:

   • For each customer, compute the hourly, daily, weekly, and monthly ride averages using Spark SQL or DataFrame operations.

3. Cohort Analysis:

   • Group customers by each week and calculate week-on-week retention.
   • Use window functions or joins to determine the number of customers returning in subsequent weeks.
   • Present the results in a cohort matrix format.