Water quality monitoring is invaluable to ensure compliance with regulations, detect trends or patterns, and advance ecological understanding. However, monitoring typically measures only a few characteristics in a small fraction of a large and complex system, and thus the information contained in monitoring data depends upon which features of the ecosystem are actually captured by the measurements. Difficulties arise when these data contain something other than intended, but this can be minimized if the purpose of the sampling is clear, and the sampling design, measurements, and data interpretations are all compatible with this purpose. The monitoring program and data interpretation must also be properly matched to the structure and functioning of the system. Obtaining this match is sometimes an iterative process that demands a close link between research and monitoring. This paper focuses on water quality monitoring that is intended to track trends in aquatic resources and advance ecological understanding. It includes examples from three monitoring programs and a simulation exercise that illustrate problems that arise when the information content of monitoring data differs from expectation. The examples show (1) how inconsistencies among, or lack of information about, the basic elements of a monitoring program (intent, design, measurement, interpretation, and the monitored system) can produce a systematic difference (bias) between monitoring measurements and sampling intent or interpretation, and (2) that bias is not just a statistical consideration, but an insidious problem that can undermine the scientific integrity of a monitoring program. Some general suggestions are provided and hopefully these examples will help those engaged in water quality monitoring to enhance and protect the value of their monitoring investment.