SQL GROUP BY Multiple Columns: Think in Result Grain

A sales manager asks:

"Show me revenue and order count by region and sales channel."

This sounds like a simple GROUP BY question.

But the important SQL question is not:

"Can I GROUP BY multiple columns?"

Yes, you can.

The better question is:

"What does one row in my result represent?"

Suppose we have this small order dataset:

DECLARE @OrderSales TABLE
(
    OrderId      INT,
    Region       VARCHAR(20),
    SalesChannel VARCHAR(20) NULL,
    Revenue      DECIMAL(10, 2)
);

INSERT INTO @OrderSales
(
    OrderId,
    Region,
    SalesChannel,
    Revenue
)
VALUES
    (101, 'North', 'Online',  120.00),
    (102, 'North', 'Online',  180.00),
    (103, 'North', 'Partner', 250.00),
    (104, 'South', 'Online',  300.00),
    (105, 'South', 'Retail',  150.00),
    (106, 'South', 'Retail',  200.00),
    (107, 'North', NULL,       90.00),
    (108, 'North', NULL,       60.00);

If the business wants one row per region, we could group only by Region.

But the requirement is more detailed:

One row per Region + SalesChannel combination.

SELECT
    Region,
    SalesChannel,
    COUNT(*) AS OrderCount,
    SUM(Revenue) AS TotalRevenue
FROM @OrderSales
GROUP BY
    Region,
    SalesChannel
ORDER BY
    Region,
    SalesChannel;

The result is:

Region   SalesChannel   OrderCount   TotalRevenue
-------  -------------  -----------  ------------
North    NULL            2            150.00
North    Online          2            300.00
North    Partner         1            250.00
South    Online          1            300.00
South    Retail          2            350.00