Analysing micro-expressions from a (failed) user testing session