Solution

To analyze the transaction behavior of bank visitors, we aim to determine how many transactions each visitor made during their visits and summarize this data to understand the distribution of transaction counts across all visits. Specifically, we want to generate a report that shows:

• The number of transactions (transactions_count) a visitor made in a single visit.
• The number of visitors (visits_count) who made that number of transactions in one visit.

This analysis helps the bank understand customer engagement and transaction patterns, which can inform service improvements and marketing strategies.

Approach Overview

1. Generate a Sequence of Possible Transaction Counts (t CTE):

   • Create a list of all possible transaction counts from 0 up to the total number of transactions to ensure that even if no visits have a certain number of transactions, the count for that transaction number is still represented.

2. Calculate Transactions per Visit (t1 CTE):

   • For each visit, count the number of transactions made by the visitor. Include visits with zero transactions by performing a LEFT JOIN between Visits and Transactions.

3. Aggregate Visits by Transaction Count:

   • Join the sequence of possible transaction counts with the actual transaction counts per visit and count how many visits correspond to each transaction count.

4. Finalize the Results:

   • Present the transactions_count alongside the corresponding visits_count, ensuring that all transaction counts from 0 to max(transactions_count) are included and ordered accordingly.

SQL Query

WITH t AS (
    SELECT Row_number() OVER() row_num
    FROM   Transactions
    UNION
    SELECT 0
),
t1 AS (
    SELECT Count(transaction_date) transaction_count
    FROM Visits v
    LEFT JOIN Transactions t
        ON v.user_id = t.user_id
           AND v.visit_date = transaction_date
    GROUP  BY v.user_id,
              v.visit_date
)
SELECT row_num AS Transactions_count,
       Count(transaction_count) visits_count
FROM   t
       LEFT JOIN t1
              ON row_num = transaction_count
GROUP  BY row_num
HAVING row_num <= (SELECT Max(transaction_count)
                   FROM   t1)
ORDER  BY row_num;

Step-by-Step Approach

Step 1: Generate the t Common Table Expression (CTE)

Create a sequence of numbers representing possible transaction counts, including 0. This ensures that even if no visits have a certain number of transactions, the count for that transaction number is still represented as 0.

SQL Snippet:

WITH t AS (
    SELECT Row_number() OVER() row_num
    FROM   Transactions
    UNION
    SELECT 0
)

Explanation:

• Row_number() OVER() row_num:
  • Assigns a unique sequential number (row_num) to each row in the Transactions table. If there are N transactions, row_num ranges from 1 to N.
• UNION SELECT 0:
  • Adds 0 to the list of row_num values to account for visits with no transactions.

Intermediate Output After Step 1 (t CTE):

+---------+
| row_num |
+---------+
| 1       |
| 2       |
| 3       |
| 4       |
| 5       |
| 6       |
| 7       |
| 8       |
| 0       |
+---------+

Step 2: Calculate Transactions per Visit (t1 CTE)

Determine the number of transactions each visitor made during each visit to the bank.

SQL Snippet:

t1 AS (
    SELECT Count(transaction_date) transaction_count
    FROM Visits v
    LEFT JOIN Transactions t
        ON v.user_id = t.user_id
           AND v.visit_date = transaction_date
    GROUP  BY v.user_id,
              v.visit_date
)

Explanation:

• LEFT JOIN Transactions t ON v.user_id = t.user_id AND v.visit_date = transaction_date:

  • Joins each visit with its corresponding transactions based on user_id and visit_date. This ensures that visits without any transactions are included with NULL values for transactions.

• COUNT(transaction_date) AS transaction_count:

  • Counts the number of transactions for each visit. Since it's a LEFT JOIN, visits with no transactions will have a transaction_count of 0.

• GROUP BY v.user_id, v.visit_date:

  • Aggregates the data for each unique visit.

Intermediate Output After Step 2 (t1 CTE):

Based on the example data:

Step 3: Aggregate Visits by Transaction Count

Determine how many visits correspond to each number of transactions (transactions_count), ensuring that all possible counts from 0 to the maximum observed are included.

SQL Snippet:

SELECT row_num AS Transactions_count,
       Count(transaction_count) visits_count
FROM   t
       LEFT JOIN t1
              ON row_num = transaction_count
GROUP  BY row_num
HAVING row_num <= (SELECT Max(transaction_count)
                   FROM   t1)
ORDER  BY row_num;

Explanation:

• LEFT JOIN t1 ON row_num = transaction_count:
  • Joins the sequence of possible transaction counts (t) with the actual transaction counts per visit (t1).
• COUNT(transaction_count) AS visits_count:
  • Counts the number of visits that have a specific transactions_count. If no visits have a certain transactions_count, the count will be 0.
• GROUP BY row_num:
  • Aggregates the data based on each transaction count.
• HAVING row_num <= (SELECT Max(transaction_count) FROM t1):
  • Ensures that only relevant transaction counts (up to the maximum observed) are included in the final result.
• ORDER BY row_num:
  • Sorts the results by transactions_count in ascending order for clarity.

Final Output:

Based on the example data, the final output will be:

+--------------------+--------------+
| Transactions_count | visits_count |
+--------------------+--------------+
|0                   |4             |
|1                   |5             |
|2                   |0             |
|3                   |1             |
+--------------------+--------------+