My first question is why does the design call for keeping aggregated line item totals at an order level? At first glance it would appear to me that you're trying to force into a table what perhaps belongs in a cube or runtime aggregating query on the server or possibly in a pivot table or reporting tool on a client. What would happen if the line items had errors one day or missing data? Once the detail problem is fixed you would have to rerun your aggregator to update that table. bleh. Without knowing more, I would say if it feels strange, it probably is. My bet is on not keeping those aggregations around in table unless you have to. (The purists would say that that is what a cube is essentially doing...storing aggregations... but that was what it was designed for.)