Multiple copies of web services reside on associated computing devices, each having an associated reputation. A client may desire to access the web service having the highest or best reputation to be ensured of a greater degree of accuracy and confidence. The client does a search, and attaches to whichever web service has the highest reputation. By running multiple copies of the web services, they may vote amongst themselves on the results in the event that one or more of the services starts giving incorrect or otherwise inconsistent results. Combining the voting with reputation data associated with each copy of the web service allows a service's reputation to be dynamically adjusted based upon how faithfully it computes the results of work items sent to it.