Using a measure as a dimension / nested aggregates.

Question

Hi,
I am working on a project for a bank, they offer customers the facility to make payments via multiple methods (Credit Card, Debit Card, Bank Transfer, Direct Debit, etc.) and they want to know how many customers use more than one payment method in a given time period (and plot that over time). Effectively their result set would look something like this:
Methods Used, No. Customers
1 , 234523
2 , 34345
3 , 978
4 , 3
In this query I have an aggregate fact including the Customer and PaymentMethod keys as well as keys for Date, Branch and other things. To answer this query in SQL is simple, I just need to use a subquery, grouping by my COUNT(DISTINCT PaymentMethod) and doing a COUNT(DISTINCT Customer), if I were to realise this in a SQL Query I get:
SELECT NoMethods,COUNT(Customer_SK) AS NoCustomers,SUM(Transactions) AS Transactions
FROM (
     SELECT Customer_SK,COUNT(DISTINCT PaymentMethod) AS NoMethods,SUM(Transactions) AS Transactions
     FROM f_Customer
     INNER JOIN d_PaymentMethod ON f_Customer.PaymentMethod_SK = d_PaymentMethod.PaymentMethod_SK
     WHERE Date_SK = 39622 AND Branch_SK = 286
     AND Customer_SK <> 0 GROUP BY Customer_SK
) x
GROUP BY NoMethods
My problem is that I don't know how to model this properly in OBIEE. The best situation I have managed to acheive is that I can call back the NoMethods and NoCustomers columns but they are pulled into Presentation Services at Customer_SK level (meaning one row per customer instead of aggregating). I managed to get it to aggregate by using a pivot table looking at one branch for one day but that's not a feasible solution as I need to run this report for all branches across months of time which would bring far too many records back to Presentation Services. In this case my NoMethods column is added into the fact table and is a COUNT DISTINCT of the PaymentMethod column from the PaymentMethod dimension, I have tried building a dimesion just for this purpose but that didn't seem to work.
If anyone has any suggestions they'd be much appreciated, thanks in advance.

Answer 1

I didn't get into detail of your particular example, but I've had a similar problem where I had to do something like AVG (AVG metric) - currently OBIEE doesn't allow for that, however, I was able to get the results I needed by applying AVG aggregation in Answers on the metric column. I know this might be too simplified, but try to give it a try. Unfortunately, the option that would work for sure for you (a workaround) would me creating a SQL view in physical layer with this pre-aggregated column, dragging it to BMM, and then selecting aggregation rule on it again in logical. Let me know if you'd like more info.
Message was edited by:
wildmight