You put the WHERE clause in the query that is more time-consuming to run -- because WHERE limits the number of records to be processed, thus improving performance.

For instance:

     SELECT a.ProductID, a.ProductDesc, b.TotalCost FROM ProductsT AS a
     INNER JOIN (
          SELECT ProductID, Sum([ProductCost]) AS TotalCost
          FROM ProductSalesT
          GROUP BY ProductID
          HAVING ProductID = "12345"
     ) AS b
     ON a.ProductID = b.ProductID

is more efficient than:

     SELECT a.ProductID, a.ProductDesc, b.TotalCost FROM ProductsT AS a
     INNER JOIN (
          SELECT ProductID, Sum([ProductCost]) AS TotalCost
          FROM ProductSalesT
          GROUP BY ProductID
     ) AS b
     ON a.ProductID = b.ProductID
     WHERE a.ProductID = "12345"

even though they (likely) will return the same result.

That's because in the former, the condition ProductID = "12345" is put inside instead of outside.  The inside query is a summation query, which is more time-consuming to perform.  Hence, putting a condition there will help the performance.

Btw, this above method uses "nested queries", which essentially let you run multiple queries with just one query.  The performance hit is not very big if the number of records being processed is kept in check.  

You can split them up into two queries, if you prefer.  But the placement of WHERE should follow the same rule -- you put it in the query that potentially retrieves the most records.