The 35C problem is a rostering problem: card data says afternoon trade does not come back until dark

The cleanest test in this dataset is the department store, because it removes the obvious explanation. Those buildings are cooled and mostly enclosed, and they still took 35% less between 3 and 6pm on days above 35C ^[4]. What heat suppresses is not time spent indoors but the trip to get there: the car park and the walk from the stop, and the decision made at half past two that it can wait. Convenience stores, reachable in a few minutes from home, held their trade ^[5]. Cooling the shop floor does not address that, and cooling the shop floor is the part of the adaptation budget most retailers have already spent.

The 7% headline is doing a lot of averaging ^[3]. One category in one three-hour window lost five times that ^[16], because the aggregate is spread across hours where the effect is small or absent. A manager rostering against the headline figure is rostering against the wrong figure.

Frequency matters as much as depth. Across the two years, 35C or above was reached on 62 days out of roughly 2,193 city-days, about 2.8% ^[2][14]. Per city per year that is around 13 or 14 days in Adelaide, ten in Melbourne, seven or eight in Sydney ^[15]. That is often enough to show up several times a summer and rare enough that nobody builds a permanent roster around it, which is why the response tends to be nothing.

Demand moves into the evening, and the evening is the part of the urban day that has been deliberately thinned out. Public transport frequency falls after the commuter rush, community centres, child care and public buildings tend to close early, and parks often lack the lighting or safety measures to be used after dark ^[7]. Libraries serve as heat refuges but shut before the temperature drops ^[13]. Seoul now keeps more than 120 public libraries open until 10pm as climate shelters ^[9], Paris and Toulouse are holding parks and swimming areas open into the night ^[10], and Dubai, where summer temperatures can pass 50C, has opened floodlit beaches overnight ^[11]. The researchers, writing at phys.org, say there is little evidence of a wider shift and that authorities in many cities are not yet planning for afternoons that are unpleasant or dangerous ^[8].

The unresolved number is the split between deferred and forgone. The authors state that some purchases were postponed and others lost altogether ^[6], without publishing the proportion ^[17]. That ratio decides the whole response. If most of the 7% is deferral, later hours recover the revenue and the only cost is scheduling labour into a different block. If a large share is lost outright, extending hours mainly redistributes trade toward whoever stays open, and the operators who move first carry staffed evening hours against sales that never existed.

Streets, building cooling and tree planting remain necessary, and the authors say so ^[12]. They are also capital projects with delivery timelines. Opening and closing times are the lever available before the next hot afternoon.