Some days I code faster than I think. As a practice, I always set nocount to on within a stored procedure to reduce the data being returned from a function call. I admit that this data is minimal and may not be worth worrying about on smaller systems but I believe in consistency and allowances for growth.
In my development database I wrote a new stored procedure to provide a quick data update on a parent / child type of table set. This procedure updates child records based on a parent ID as well as a child record seniority. I tested the procedure by running it and pulling the child records for my given parent ID. However, I had missed a portion of the where clause.
Do you see the flaw? I had actually updated the entire child record table but only reviewed the data for a single parent ID. Setting nocount to on masked this flaw and I didn't catch it until a code review later in the day.
To sum up: using nocount is one of many methods to reduce your data stream, but make sure you turn it off during development and testing.