My Technique is 20% Faster: Problems with Reports of Speed Improvements in HCI
Pierre Dragicevic · HAL (Le Centre pour la Communication Scientifique Directe) · 2012
In human-computer interaction (HCI), speed improvements are often used as an informal measure of effect size, with statements such as "technique A was about 20% faster than technique B". Such statements are however surrounded by confusion. A quick analysis of the literature reveals that half of the time, the term "% faster" actually refers to a percent change in task completion time, while one third of the time it correctly refers to a percent change in speed. The rest of the time, the meaning is unknown or the calculation is wrong. We explain why percent changes are inherently confusing, and propose to focus effect size discussions on ratios, or on measures of percent difference where the normalizer is the mean of the compared values. When percent changes need to be reported, it is suggested to focus the discussion on improvements in task completion times rather than speed.