Creating averages and detecting increases higher than x% in SQL - sql

Question

I want to create the following in SQL Server 2012: (I've found that the best way to explain it is with tables).
I have the date of purchase, the customer id and the price the customer paid in a table like this:
DateOnly Customer Price
2012/01/01 1 50
2012/01/01 2 60
2012/01/01 3 80
2012/01/02 4 40
2012/01/02 5 30
2012/01/02 1 55
2012/01/03 6 80
2012/01/04 2 90
What I need to do then is to keep a register of the average price paid by a customer. Which would be as follows:
DateOnly Customer Price AveragePrice
2012/01/01 1 50 50
2012/01/01 2 60 60
2012/01/01 3 80 80
2012/01/02 4 40 40
2012/01/02 5 30 30
2012/01/02 1 55 52.5
2012/01/03 6 80 80
2012/01/04 2 90 75
And finally, I need to select the rows which have caused an increase higher than 10% in the averageprice paid by a customer.
In this case, the second order of customer 2 should be the only one to be selected, as it introduced an increase higher than 10% in the average price paid by this customer.
Hence, the resulting table should be as follows:
DateOnly Customer Price AveragePrice
2012/01/04 2 90 75
Thanks in advance for your help.

Answer 1

First CTE is to prepare your data = assign row_numbers to each customer's purchase, to be used in joins further.
Second CTE is recursive and it does all the work in process. First part is to get each customer's first purchase and recursive part joins on next purchase and calculates TotalPrice, AveragePrice and Increase.
At the end just select the rows with Increase more than 10%.
WITH CTE_Prep AS
(
SELECT *, ROW_NUMBER() OVER (PARTITION BY Customer ORDER BY DateOnly) RN
FROM Table1
)
,CTE_Calc AS
(
SELECT *, Price AS TotalPrice, CAST(Price AS DECIMAL(18,2)) AS AveragePrice, CAST (0 AS DECIMAL(18,2)) AS Increase
FROM CTE_Prep WHERE RN = 1
UNION ALL
SELECT p.*
, c.TotalPrice + p.Price AS TotalPrice
, CAST(CAST(c.TotalPrice + p.Price AS DECIMAL(18,2)) / p.RN AS DECIMAL(18,2)) AS AveragePrice
, CAST(CAST(CAST(c.TotalPrice + p.Price AS DECIMAL(18,2)) / p.RN AS DECIMAL(18,2)) / c.AveragePrice AS DECIMAL(18,2)) AS Increase
FROM CTE_Calc c
INNER JOIN CTE_Prep p ON c.RN + 1 = p.RN AND p.Customer = c.Customer
)
SELECT * FROM CTE_Calc
WHERE Increase > 1.10
SQLFiddle DEMO

Answer 2

Interesting problem.
You can get the average without the current purchase by subtracting the price on each row from the sum of all prices for the row. This observation -- in combination with window functions -- gives the information needed to get the rows you are looking for:
select *
from (select t.*,
avg(price) over (partition by customer) as avgprice,
sum(price) over (partition by customer) as sumprice,
count(price) over (partition by customer) as cntprice
from table1 t
) t
where (case when cntprice > 1
then (sumprice - price) / (cntprice - 1)
end) > avgprice*1.1;
Note the use of the case in the where clause. There is a potential divide by zero problem. SQL Server guarantees that the when part of the case will be evaluated before the then part (in the situation). So this is safe from that problem.