A big challenge for smart grid application scenarios will be handling the massive amount of data that is expected to be collected from data sources such as smart meters and sent through the communication backhaul to the grid operator. One approach to reduce data volume given some application sampling rate is to aggregate the data stream within the network itself. While data aggregation along the communications network can reduce data volume and alleviate congestion issues, it can also reduce the granularity of application data reaching the utility control center, thus being not as useful. This work explores the tradeoff between the quality of application data collected and communications network congestion. It defines the concept of data granularity and presents two case studies on the impact of data granularity on application quality/accuracy. Finally, an algorithm is proposed and evaluated that demonstrates how the data granularity and application quality tradeoff can be managed at intermediate relay nodes such as data concentrators.