I'm running a use test for 12 participants split into 2 sets of 3 subgroups (user type and user location - these are NOT mutually exclusive groups).
I need to produce 7 different SUS results: overall, u-type-a, u-type-b, u-type-c, geo-loc-a, geo-loc-b, geo-loc-c.
Obviously, SUS relies on self-reported statistics given after the experience - this means that the results are not necessarily 100% accurate and could probably benefit from being presented with some sort of margin for error (for error read 'misremembering', 'misunderstanding', 'alternate system bias', etc).
As the SUS is a relatively complex calculation there are plenty of points to add in an error margin: on the question scores, on the score total, on the percentage per user, or on the average for the user group.
My question is this: How much margin, if any, should I allow for and where in the calculation should I add it?