If you use an ordinary TQuery component as a decision dataset, you must manually set up the SQL statement, taking care to supply a GROUP BY phrase which contains the same fields (and in the same order) as the SELECT phrase. 

The SQL should look similar to this:

SELECT ORDERS."Terms", ORDERS."ShipVIA",
  ORDERS."PaymentMethod", SUM( ORDERS."AmountPaid" )
FROM "ORDERS.DB" ORDERS
GROUP BY ORDERS."Terms", ORDERS."ShipVIA", ORDERS."PaymentMethod"

The ordering of the SELECT fields should match the ordering of the GROUP BY fields. Queries are described in more detail in Using TQuery. 

With TTable, you must supply information to the decision cube about which of the fields in the query are grouping fields, and which are summaries. To do this, Fill in the Dimension Type for each field in the DimensionMap of the Decision Cube. You must indicate whether each field is a dimension or a summary, and if a summary, you must provide the summary type. Since pivoting averages depend on SUM/COUNT calculations, you must also provide the base field name to allow the decision cube to match pairs of SUM and COUNT aggregators.