Duplication involving SUM, LEFT JOIN and GROUP BY

Question

I have a problem involving a SUM, LEFT OUTER JOIN and GROUP BY commands, but can\'t figure out where my error is.

I have two tables, one for customer transactions an

Answer 1

The query will be counting the transaction amount twice for 2007 due to the fact you have two claims, so the transaction amount will be counted twice.

i.e. the returned data being used is:

Customer | Transaction Year | Transaction Amount | Claim Amount
----------------------------------------------------------------
A        | 2007             | 100                | 30
A        | 2007             | 100                | 40
A        | 2008             | 80                 |
A        | 2008             | 50                 |
A        | 2009             | 210                | 110

Something like the following, although not pretty, should resolve the problem:

SELECT 
   t.Customer
   ,t.Year
   ,[Transaction Amount] = SUM(t.[Transaction Amount])
   ,[Claim Amount] = c.[Claim Amount]
FROM 
    Transactions t
    LEFT JOIN (
        SELECT 
            Customer
            ,Year
            ,SUM([Claim Amount])
        FROM
           Claims
        GROUP BY
           Customer, Year
    ) c ON c.Customer = t.Customer c.Year = t.Year
GROUP BY t.Customer, t.Year, c.[Claim Amount]

Answer 2

With T as (
    SELECT  Customer,
            [Transaction Year],
            sum(Amount) AS TransactionAmount
    FROM Transactions
    GROUP BY Customer, [Transaction Year]
), C AS 
    SELECT  Customer,
            [Claim Year],
            sum(Amount) as ClaimAmount
    FROM Claims
    GROUP BY Customer, [Claim Year]
)
SELECT  T.Customer,
        [Transactions Year],
        TransactionAmount,
        ClaimAmount
FROM    T
   LEFT JOIN C ON C.Customer = T.Customer
      AND [Transactions Year] = [Claim Year]

Answer 3

So the first step to see what is happening is to remove the SUMs and just select the transaction amount and claim amount. That way you can see what data is being returned. You'll see that the join on A/2007 will have the transaction amount twice, since it's joining each row to the claims table.

One solution is to use subquerys, like you said, to do the SUMs separately prior to joining.

SELECT 
   Transactions.Customer,
   Transactions.Year,
   SumTransaction,
   SumClaim
FROM (
      select Customer, Year, sum(Transaction Amount) SumTransaction 
      from Transactions
      group by Customer, Year
   ) Transactions
   LEFT JOIN (
      select Customer, Year, sum(Claim Amount) sumClaim 
      from Claims
      group by Customer, Year
   ) Claims
   ON Claims.Customer = Transactions.Customer
      AND Transactions.Year = Claims.Year

Another possible solution given your restrictions:

SELECT 
   Transactions.Customer,
   Transactions.Year,
   SUM(Transaction Amount),
   (SELECT SUM(Claim Amount) from Claims where Claims.Customer = Transactions.Customer and Claims.Year = Transactions.Year)
FROM 
   Transactions
GROUP BY
   Customer, Year

Third possible solution!! This one does not require any subqueries! See this SQL Fiddle

select
    t.Customer,
    t.Year,
    sum(distinct t.Amount),
    sum(c.Amount)
from
    Transactions t
    left join Claims c
        on  t.Customer = c.Customer
            and t.Year = c.year
group by
    t.Customer,
    t.Year