Sql query group by question, creating monthly averages - sql

Question

I am trying to compile a table which shows three columns:
product name, average monthly sales volume and average monthly sales price for said product.
I am using adventureworks 2019.
I have written the below query based upon some help I previously received. I have summed the order quantity and unit price for each product and then in the outer query averaged them.
SELECT
Product_Name,
AVG(Sales_Volume) AS Avg_Sales_Volume,
AVG(Price) AS Avg_Price
FROM
(SELECT
PP.[Name] AS Product_Name,
SUM(SSOD.[OrderQty]) AS Sales_Volume,
SUM(SSOD.[UnitPrice]) AS Price,
FORMAT(SSOH.[OrderDate], 'MM-yyyy') AS Month_Year
FROM
[Sales].[SalesOrderHeader] AS SSOH
LEFT JOIN Sales.SalesOrderDetail AS SSOD
ON SSOH.SalesOrderID = SSOD.SalesOrderID
LEFT JOIN production.product AS PP
ON SSOD.ProductID = PP.ProductID
GROUP BY PP.[Name], SSOH.[OrderDate]) AS T
--WHERE Product_Name = 'Road-150 Red, 44' (doing this for reference)
GROUP BY T.Product_Name, Month-Year
If I group by Month-Year I get around 3600 rows, where as if I don't, I get 266. I am confused about this. Basically, I am not sure what it is actually showing.
To some of you this may seem really basic, but it feels like I cannot get my head around it.
Could anyone take a bit of time to explain this to me?
Thanks
Product_Name Avg_Sales_Volume Avg_Price
LL Mountain Frame - Silver, 48 14 844.96
LL Touring Frame - Blue, 50 26 2100.546
Women's Mountain Shorts, L 13 161.494
Road-550-W Yellow, 44 3 1890.7332
HL Road Frame - Red, 48 18 6025.3137
Mountain-500 Silver, 42 8 1395.0116
(266 rows)
Query with group by Month-Year also:
Product_Name Avg_Sales_Volume Avg_Price
Road-150 Red, 44 1 6758.9544
HL Mountain Frame - Silver, 46 15 4465.6362
AWC Logo Cap 14 76.4672
Long-Sleeve Logo Jersey, L 2 102.611
Road-150 Red, 56 2 6817.546
Mountain-500 Silver, 52 13 2118.7125
LL Touring Frame - Yellow, 62 62 5001.30
ML Mountain Frame-W - Silver, 40 115 6546.3382
(3862 rows)

Answer 1

A few notes first, then an explanation of what you're seeing.
First, your outer query has GROUP BY Month-Year with a dash, but I believe this is intended to be the column Month_Year (with an underscore) from the subquery.
Next, I like your use of AS to explicitly define aliases on columns, as well as your use of square brackets [ and ] around object names rather than quoted identifiers.
Finally, start getting in the practice of using schema- and alias-identifiers throughout your code - especially when working with subqueries. They'll make it much more clear which objects should be returned, and will prevent "ambiguous column" errors.
Explanation:
The difference in row-counts returned for each of your two queries is due to how those queries instruct the grouping to occur.
GROUP BY T.Product_Name; returns 266 rows because it is aggregating all of the sales volume and pricing information for the given product across all time. Changing up your subquery a bit to read:
SELECT COUNT(DISTINCT pp.Name)
FROM [Sales].[SalesOrderHeader] AS SSOH
LEFT JOIN Sales.SalesOrderDetail AS SSOD
ON SSOH.SalesOrderID = SSOD.SalesOrderID
LEFT JOIN production.product AS PP
ON SSOD.ProductID = PP.ProductID
Results in 266 rows being returned, indicating that there are 266 distinct product names included in that set.
When you add more grouping conditions (such as T.Month_Year), you are telling the engine to make "subgroups" in the aggregation structure.
GROUP BY T.Product_Name, T.Month_Year runs your aggregates for each of those distinct 266 product names as well as aggregating the data for each distinct T.Month_Year value that appears in each of your 266 product name groups.
Examining those row-counts a bit closer, the two-condition grouping returns 3,862 rows while the single-condition grouping returns 266 rows. Across those 3,862 rows there are 266 distinct product names represented an average of 14.52 times (3862 / 266 = 15.518...). If you assumed that every product had at least one sale per month, then you might conclude that we are looking at slightly more than one year's worth of sales data here. More likely, this is a set of several years of sales data with a lot of variation in sales volume and frequency between products.
ADDENDUM: Adding your GROUP BY columns to the SELECT illustrates the difference in result sets:
SELECT Product_Name,
T.Month_Year,
AVG(Sales_Volume) AS Avg_Sales_Volume,
AVG(Price) AS Avg_Price
FROM (SELECT PP.[Name] AS Product_Name,
SUM(SSOD.[OrderQty]) AS Sales_Volume,
SUM(SSOD.[UnitPrice]) AS Price,
FORMAT(SSOH.[OrderDate], 'MM-yyyy') AS Month_Year
FROM [Sales].[SalesOrderHeader] AS SSOH
LEFT JOIN Sales.SalesOrderDetail AS SSOD
ON SSOH.SalesOrderID = SSOD.SalesOrderID
LEFT JOIN production.product AS PP
ON SSOD.ProductID = PP.ProductID
GROUP BY PP.[Name], SSOH.[OrderDate]) AS T
--WHERE Product_Name = 'Road-150 Red, 44' (doing this for reference)
GROUP BY T.Product_Name,
T.Month_Year
ORDER BY Product_Name;
Examining the results shows that each product name also has records for any month in which that product sold:
+--------------------------+-----------+-------+-----------+
| ProductName | Month_Year|Avg_Vol| Avg_Price |
+--------------------------+-----------+-------+-----------+
| All-Purpose Bike Stand | 12-2013 | 1 | 193.0714 |
| All-Purpose Bike Stand | 06-2014 | 1 | 218.625 |
| All-Purpose Bike Stand | 05-2014 | 1 | 187.909 |
| All-Purpose Bike Stand | 10-2013 | 1 | 212.00 |
| AWC Logo Cap | 02-2014 | 6 | 57.7928 |
| AWC Logo Cap | 02-2012 | 48 | 93.357 |
| AWC Logo Cap | 08-2011 | 68 | 103.73 |
| AWC Logo Cap | 01-2013 | 124 | 129.4896 |
| AWC Logo Cap | 03-2014 | 21 | 71.1747 |
+--------------------------+-----------+-------+-----------+