I want to pull out top 3 selling products for different product category per tag.
Data looks like this:
tag | product_name | product_category | order_count
tag1 | product1 | category1 | 100
tag1 | product2 | category2 | 80
tag1 | product3 | category2 | 60
tag1 | product4 | category3 | 50
......
I know how to pull out top 3 selling products per tag using ROW_NUMBER(), but it will return product1,product2,product3. I don't want product3 because it belongs to the same category as product2. I want product4 instead. How to do this in SQL server?
First ROW_NUMBER removes duplicate rows per tag and product_category, second ROW_NUMBER selects top 3 selling products per tag
;WITH cte AS
(SELECT *, ROW_NUMBER() OVER(PARTITION BY tag, product_category ORDER BY order_count DESC) AS rn
FROM yourtable
), cte2 AS
(SELECT *, ROW_NUMBER() OVER(PARTITION BY tag ORDER BY order_count DESC) AS rn2
FROM cte
WHERE rn = 1
)
SELECT *
FROM cte2
WHERE rn2 <= 3
Demo on SQLFiddle
Next one uses a derived table
;WITH cte AS
(SELECT t2.tag, t2.product_name, t2.product_category, t2.order_count,
ROW_NUMBER() OVER(PARTITION BY t2.tag ORDER BY order_count DESC) AS rn
FROM (SELECT tag, product_category, MAX(order_count) AS maxCount
FROM yourtable
GROUP BY tag, product_category
) t1 JOIN yourtable t2 ON t1.tag = t2.tag
AND t1.product_category = t2.product_category
AND maxCount = order_count
)
SELECT *
FROM cte
WHERE rn <= 3
Demo on SQLFiddle
You can use RANK()
(or ROW_NUMBER()
) as long as you use PARTITION BY
. This in combination with TOP()
should work well assuming you're using SQL Server 2005+:
with cte as (
select tag,
product_name,
product_category,
order_count,
rank() over (partition by product_category
order by product_category, order_count desc) rnk
from yourtable
)
select top 3 tag, product_name, product_category, order_count
from cte
where rnk = 1
order by order_count desc
This will produce the following:
TAG PRODUCT_NAME PRODUCT_CATEGORY ORDER_COUNT
tag1 product1 category1 100
tag1 product2 category2 80
tag1 product4 category3 50
I would suggest doing it with one select only
declare @t table (
tag varchar(10),
product_name varchar(10),
product_category varchar(10),
order_count int
);
insert into @t values
('tag1', 'product1', 'category1', 100),
('tag1', 'product2', 'category2', 80 ),
('tag1', 'product3', 'category2', 60 ),
('tag1', 'product4', 'category3', 50 ),
('tag1', 'product5', 'category4', 40 );
select top 3
*
from
@t
order by
row_number() over(partition by product_category order by order_count desc),
order_count desc;
SQL Fiddle