A new paper from the School of Continuous Improvement tests a building's fire-drill assembly headcount against its own swipe-card log, finding the two agree only moderately and that a marshal refresher briefing doesn't close the gap.
A term of matched fire-drill tallies and swipe-card logs agree on six of the building's nine floors, and diverge, in different directions, on the rest
When a fire alarm sounds in the building, the number that reaches Facilities and Safety first isn’t the one from the door: it’s the headcount marshals radio in from the lawn, the forecourt, and the car park below. A new paper from the School of Continuous Improvement asked how closely that number tracks what the building’s own swipe-card system would say for the same alarm, and found the two agree on some floors, and for different reasons on the ones where they don’t.
A number’s job doesn’t end once it’s accurate enough to act on. Ours told us that within the hour, and we only checked what it actually was afterwards.
— Dr Mirela Hanke, Postdoctoral Fellow, School of Continuous Improvement
Over one teaching term, Dr Hanke, Dr Renke Sabel, and Dr Osei Vandermeer matched every fire-drill assembly headcount at a nine-storey administration building against what the same building’s swipe-card system recorded in the minutes before each alarm, floor by floor, then tested whether an added marshal briefing partway through the term closed any gap between the two. It didn’t: the lowest floors’ tallies ran high, the highest floors’ ran low, and the pattern held before and after the briefing.
Dr Sabel called the finding “a fairly ordinary thing for an indicator to do once nobody’s asked it to prove itself in a while,” and said the School is weighing whether the roster formula that leans on the headcount deserves a standing check of its own.
The School reads the result as evidence that even its most habitual instruments are worth an occasional second look, particularly the ones nobody remembers choosing over an available alternative.
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.hq8wqq.