This video provides a walkthrough of the challenge requirements for segmenting a dataset by two dimensions (such as acquisition month and user region) and calculating advanced retention metrics. You will see how to use pandas to group data, create multi-level cohorts, and compute retention and churn rates for each segment.

To complete this challenge, follow these steps:

- Use a **pandas DataFrame** containing user activity data, with columns such as `user_id`, `acquisition_month`, `region`, and `activity_month`;
- Segment the users by both `acquisition_month` and `region` to create multi-level cohorts;
- For each cohort, calculate the number of users retained in each subsequent month after acquisition;
- Compute the **retention rate** for each cohort as the percentage of users active in a given month compared to the original cohort size;
- Calculate the **churn rate** as 1 minus the retention rate for each period.

You will need to use **pandas** grouping and aggregation methods to perform these calculations efficiently.

import unittest
import user_code
import importlib
import pandas as pd

class TestTask(unittest.TestCase):
    def setUp(self):
        self.data = {
            'user_id': [1, 1, 2, 2, 3, 4, 4, 5, 6, 6],
            'acquisition_month': [1, 1, 1, 1, 2, 2, 2, 2, 2, 2],
            'activity_month': [1, 2, 1, 2, 2, 2, 3, 2, 2, 3]
        }
        self.df = pd.DataFrame(self.data)

    def test_returns_dataframe_and_columns(self):
        import user_code
        importlib.reload(user_code)
        result = user_code.calculate_retention_by_acquisition(self.df)
        expected_columns = ['cohort', 'months_since_acquisition', 'retention_rate']
        _dynamic_test(
            self,
            isinstance(result, pd.DataFrame),
            "Returned object is a pandas DataFrame.",
            f"Expected DataFrame, got {type(result)}."
        )
        _dynamic_test(
            self,
            all(col in result.columns for col in expected_columns),
            f"Result contains all required columns: {expected_columns}.",
            f"Missing columns in result. Got columns: {list(result.columns)}."
        )

    def test_cohort_definition(self):
        import user_code
        importlib.reload(user_code)
        result = user_code.calculate_retention_by_acquisition(self.df)
        cohort_values = set(result['cohort'])
        expected_cohorts = set([1, 2])
        _dynamic_test(
            self,
            cohort_values == expected_cohorts,
            f"Cohorts are defined as acquisition_month: {expected_cohorts}",
            f"Expected cohorts: {expected_cohorts}, got: {cohort_values}"
        )

    def test_retention_rate_calculation(self):
        import user_code
        importlib.reload(user_code)
        result = user_code.calculate_retention_by_acquisition(self.df)
        # For cohort 1, months_since_acquisition=0: users 1 and 2
        # cohort size: users 1 and 2, so 2
        # active users at month 0: 2
        # retention rate: 2/2 = 1.0
        row = result[(result['cohort'] == 1) & (result['months_since_acquisition'] == 0)]
        rr = float(row['retention_rate'].iloc[0]) if not row.empty else None
        _dynamic_test(
            self,
            rr is not None and abs(rr - 1.0) < 1e-8,
            "Retention rate for cohort 1 at month 0 is correct (1.0)",
            f"Expected 1.0, got {rr}"
        )
        # For cohort 1, months_since_acquisition=1: users 1 and 2 again, so 2/2 = 1.0
        row = result[(result['cohort'] == 1) & (result['months_since_acquisition'] == 1)]
        rr = float(row['retention_rate'].iloc[0]) if not row.empty else None
        _dynamic_test(
            self,
            rr is not None and abs(rr - 1.0) < 1e-8,
            "Retention rate for cohort 1 at month 1 is correct (1.0)",
            f"Expected 1.0, got {rr}"
        )
        # For cohort 2, months_since_acquisition=0: users 3, 4, 5, 6
        # cohort size: users 3, 4, 5, 6, so 4
        # active users at month 0: 4
        # retention rate: 4/4 = 1.0
        row = result[(result['cohort'] == 2) & (result['months_since_acquisition'] == 0)]
        rr = float(row['retention_rate'].iloc[0]) if not row.empty else None
        _dynamic_test(
            self,
            rr is not None and abs(rr - 1.0) < 1e-8,
            "Retention rate for cohort 2 at month 0 is correct (1.0)",
            f"Expected 1.0, got {rr}"
        )
        # For cohort 2, months_since_acquisition=1: users 4, 6, so 2/4 = 0.5
        row = result[(result['cohort'] == 2) & (result['months_since_acquisition'] == 1)]
        rr = float(row['retention_rate'].iloc[0]) if not row.empty else None
        _dynamic_test(
            self,
            rr is not None and abs(rr - 0.5) < 1e-8,
            "Retention rate for cohort 2 at month 1 is correct (0.5)",
            f"Expected 0.5, got {rr}"
        )

def _dynamic_test(test_case, condition, success_message, failure_message):
    if condition:
        test_case._testMethodName = success_message
        test_case.assertTrue(True, success_message)
    else:
        test_case._testMethodName = failure_message
        test_case.fail(failure_message)


test_main.py

Master cohort analysis techniques using Python. This course covers cohort data structuring, advanced segmentation, retention analytics, and actionable business insights, with hands-on challenges in each section.

Learn how to structure, preprocess, and prepare data for advanced cohort analysis, including cohort assignment and feature engineering.

Dive into advanced segmentation strategies and retention metric calculations to uncover deeper insights from cohort data.

Visualize cohort data and extract actionable business insights using advanced plotting and interpretation techniques.

Challenge: Advanced Segmentation and Retention

Solution