Summary  
This chapter shows how to use window functions (e.g., LAG, LEAD) combined with date functions (e.g., DATE_TRUNC, EXTRACT) to compute period-over-period metrics like month-over-month and year-over-year by aggregating and comparing values across consistent time intervals.

General domain of usage  
Business analytics

Period-over-period analysis is a core technique in business analytics, allowing you to measure **growth**, **trends**, and **changes** by comparing metrics across consistent time intervals. Common period-over-period metrics include **month-over-month (MoM)** and **year-over-year (YoY)** comparisons. These analyses help you quickly identify whether performance is improving, declining, or stable.

SQL provides powerful tools for these comparisons, especially through the use of window functions like `LAG` and `LEAD`, which enable you to reference values from previous rows in a partitioned and ordered set. Date functions such as `DATE_TRUNC`, `EXTRACT`, and arithmetic operations on dates allow you to group, filter, and align data by specific periods, making it possible to compare like-for-like intervals. By combining window functions with date logic, you can efficiently calculate changes between periods directly in your queries.

- **MoM (Month-over-Month):** Measures the percentage or absolute change in a metric from one month to the next;  
- **YoY (Year-over-Year):** Compares the same metric for a specific period (such as a month or quarter) to the same period in the previous year, revealing seasonal trends and annual growth.

Definition

SELECT
  DATE_TRUNC('month', order_date) AS month,
  SUM(total) AS monthly_sales,
  LAG(SUM(total)) OVER (ORDER BY DATE_TRUNC('month', order_date)) AS last_month_sales,
  SUM(total) - LAG(SUM(total)) OVER (ORDER BY DATE_TRUNC('month', order_date)) AS sales_difference
FROM orders
GROUP BY DATE_TRUNC('month', order_date)
ORDER BY month;

The marketing team often requests a report that tracks **monthly active users** and highlights their **MoM growth rate**. This kind of analysis helps you spot engagement trends, evaluate the impact of campaigns, and plan future initiatives more effectively.

SELECT
  DATE_TRUNC('month', signup_date) AS month,
  COUNT(DISTINCT user_id) AS active_users,
  LAG(COUNT(DISTINCT user_id)) OVER (ORDER BY DATE_TRUNC('month', signup_date)) AS last_month_active,
  ROUND(
    100.0 * (COUNT(DISTINCT user_id) - LAG(COUNT(DISTINCT user_id)) OVER (ORDER BY DATE_TRUNC('month', signup_date)))
    / NULLIF(LAG(COUNT(DISTINCT user_id)) OVER (ORDER BY DATE_TRUNC('month', signup_date)), 0),
    2
  ) AS mom_growth_percent
FROM users
WHERE is_active = TRUE
GROUP BY DATE_TRUNC('month', signup_date)
ORDER BY month;

import unittest
import psycopg2
import pandas as pd
import re

class TestPostgresQuery(unittest.TestCase):
    @classmethod
    def setUpClass(cls):
        cls.connection = psycopg2.connect(
            host="localhost",
            user="agent",
            password="qwerty",
            dbname="testdb",
            sslmode="disable"
        )
        cls.cursor = cls.connection.cursor()

    def test_columns_present(self):
        with open("query.sql", "r") as file:
            query = file.read()
        self.cursor.execute(query)
        columns = [desc[0] for desc in self.cursor.description]
        required_cols = {"category", "year", "yearly_revenue", "last_year_revenue", "revenue_change"}
        condition = required_cols.issubset(set([normalize_text(c) for c in columns]))
        _dynamic_test(
            self,
            condition,
            "All required columns are present.",
            f"Missing required columns: {required_cols - set([normalize_text(c) for c in columns])}"
        )

    def test_one_row_per_category_per_year(self):
        with open("query.sql", "r") as file:
            query = file.read()
        # Solution query for reference
        solution_query = '''
SELECT category, EXTRACT(YEAR FROM sale_date) AS year, SUM(total_sales) AS yearly_revenue
FROM product_sales
GROUP BY category, EXTRACT(YEAR FROM sale_date)
ORDER BY category, year;
'''
        self.cursor.execute(solution_query)
        solution_rows = self.cursor.fetchall()
        self.cursor.execute(query)
        user_rows = self.cursor.fetchall()
        # Compare number of rows
        condition = len(user_rows) == len(solution_rows)
        _dynamic_test(
            self,
            condition,
            "Returned one row per category per year.",
            f"Expected {len(solution_rows)} rows, got {len(user_rows)} rows."
        )

    def test_yearly_revenue_correct(self):
        with open("query.sql", "r") as file:
            query = file.read()
        solution_query = '''
SELECT category, EXTRACT(YEAR FROM sale_date) AS year, SUM(total_sales) AS yearly_revenue
FROM product_sales
GROUP BY category, EXTRACT(YEAR FROM sale_date)
ORDER BY category, year;
'''
        self.cursor.execute(solution_query)
        sol = self.cursor.fetchall()
        sol_df = pd.DataFrame(sol, columns=["category", "year", "yearly_revenue"])
        self.cursor.execute(query)
        user = self.cursor.fetchall()
        user_cols = [desc[0] for desc in self.cursor.description]
        user_df = pd.DataFrame(user, columns=user_cols)
        # Compare values for (category, year) pairs
        merged = pd.merge(sol_df, user_df, how='left', left_on=["category", "year"], right_on=["category", "year"])
        # Compare yearly_revenue
        mismatches = merged[abs(merged["yearly_revenue_x"] - merged["yearly_revenue_y"]) > 0.01]
        condition = mismatches.empty
        _dynamic_test(
            self,
            condition,
            "yearly_revenue is correct for each category/year.",
            f"Incorrect yearly_revenue for rows: {mismatches[['category','year','yearly_revenue_x','yearly_revenue_y']].to_dict(orient='records')}"
        )

    def test_last_year_revenue_and_revenue_change(self):
        with open("query.sql", "r") as file:
            query = file.read()
        # Run user query
        self.cursor.execute(query)
        user = self.cursor.fetchall()
        user_cols = [desc[0] for desc in self.cursor.description]
        user_df = pd.DataFrame(user, columns=user_cols)
        # Sort for lag comparison
        user_df = user_df.sort_values(["category", "year"])
        # Check lag and difference
        errors = []
        for cat in user_df["category"].unique():
            df_cat = user_df[user_df["category"] == cat].sort_values("year")
            prev_rev = None
            for idx, row in df_cat.iterrows():
                curr_rev = row["yearly_revenue"]
                last_year = row["last_year_revenue"]
                diff = row["revenue_change"]
                if pd.isnull(prev_rev):
                    if not pd.isnull(last_year):
                        errors.append({"category": cat, "year": row["year"], "msg": "last_year_revenue should be NULL for first year"})
                    if not pd.isnull(diff):
                        errors.append({"category": cat, "year": row["year"], "msg": "revenue_change should be NULL for first year"})
                else:
                    # Both should be floats
                    if abs(last_year - prev_rev) > 0.01:
                        errors.append({"category": cat, "year": row["year"], "msg": f"last_year_revenue incorrect: {last_year} vs {prev_rev}"})
                    if abs(diff - (curr_rev - prev_rev)) > 0.01:
                        errors.append({"category": cat, "year": row["year"], "msg": f"revenue_change incorrect: {diff} vs {curr_rev - prev_rev}"})
                prev_rev = curr_rev
        condition = len(errors) == 0
        fail_msg = "\\n".join([f"{e['category']} {e['year']}: {e['msg']}" for e in errors])
        _dynamic_test(
            self,
            condition,
            "last_year_revenue and revenue_change are correct for all rows.",
            fail_msg
        )

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)

def normalize_text(text):
    text = text.lower()
    text = re.sub(r"\s{2,}", " ", text)
    text = re.sub(r"\s*([,:?])\s*", r"\1 ", text)
    return text.strip()

if __name__ == "__main__":
    unittest.main()


test_solution.py

Master advanced SQL techniques for building analytical reports. Learn to solve real-world business problems using multi-level aggregations, window functions, cohort and funnel analysis, KPI calculations, and more. Each chapter features theory, a realistic business use case, and a hands-on coding task.

Explore advanced SQL concepts and apply them to real-world analytical reporting scenarios across multiple business domains.

Period-over-Period Comparisons (MoM, YoY)

Lösung