Assessment tests Will involve full prototypes, will focus on tasks, may involve quantitative measurements Validative tests Objectives and goals are typically quantitative Results are used to set standards, performance criteria Test involves reduced interaction between test participant and moderator Test conditions need to be fairly well controlled
You can also test how many features the user can remember during a debriefing session. The problem with metrics-based testing is that the individual users may vary widely -- measurements of speed or memory or whatever may depend more on the characteristics of a particular user than of the site or interface you are testing. One solution is to test so many users that you can apply statistical test of validity to the results. This can obviously be expensive. It also requires some statistical expertise.