The team, writing in the journal Scientific Research, worked with a call centre to analyse data from 3013 calls: the true call duration and whether the customer felt they received (using a simple yes/no response in each case) good service, sufficient information, and, the key measure, a timely service, termed time satisfaction. An initial analysis showed time satisfaction was correlated quite highly with satisfaction with service and information, and a more modest negative correlation with call duration. The size of the effects weren't directly compared.
A follow-up analysis split the data into four groups based on actual call time; for instance the 'low' group contained calls under two minutes in duration. In every group, a 'yes' for time satisfaction was much more likely to be found alongside yeses for service and information. Meanwhile, the relationship between time satisfaction and actual time was much milder, and in the low time group the effect was too weak to be significant. The authors argue that 'with waiting times being so low, time lost its value and that satisfaction with information and service were more important'.
But wait. Let's imagine that data had been split by call time, but rather than four groups there were many; so many that a single one only contained calls lasting 10m30s to 10m31s. We'd be unsurprised if within that group (and all its counterparts), call time was irrelevant; the range of possible times is so restricted that there is no interesting difference left. The chosen analysis produces a milder version of this 'restriction of range', disproportionately reducing the chances of detecting any effects for actual call time.
The study shows that satisfaction on timely call handling is coloured by factors aside from actual call time, and it's good to remind organisations that perceptions are not formed solely by such objective features. However, this research design doesn't actually put us in a position to rank these different aspects of a call.