It's very cumbersome to look at the dashboard, click on the link to a given build's info, click the build log, find which step failed. This process is exacerbated in our deployment because we use a master/slave approach with all builds scheduled through a single instance which basically does a source checkout, publish the source, trigger remote builds via QuickBuild repository, then any local cleanup (mostly publishing collecting and publishing build artifacts). The vast majority of the failed steps reported on this machine are the QB checkouts which do the actual builds. We publish those logs so we have them available to browse on the master server, but we still need to browse the build_log to find out which platform build log to look at. So having a dashboard of the individual steps, would make a much more efficient build failure analysis.

Another facet that would tie in nicely to a per-step dashboard is if the output from each step was logged separately, so that the log from each step could be browsed from the per-step dashboard. At the end of the build the logs from each step could just be concatenated into a single whole-build log.