Event-level modelling of showering behaviour in German households: Triangulating smart-meter, in-situ survey and contextual public data

Abstract

Residential water demand is strongly shaped by showering behaviour, yet quantifying drivers at the event level remains rare. We first explored and identified key patterns across 8550 sensor-logged showering events from 319 German households collected during winter. We then combined this high-resolution smart-meter data with time-stamped post-shower surveys to analyse shampoo-shower duration. The Gamma mixed model (14 predictors spanning demographics, context, in-shower behaviour and hair characteristics) explained 68.4% of deviance (AIC = 4281.6). Behavioural predictors showed the clearest associations: showers with two or more shampoo applications were 27% longer than those with one, while showers with three or more water pauses were 28% longer than those with one. Some contextual and demographic predictors also mattered: soft water areas were linked to 23% longer durations, and younger/mid-age adults had durations up to 31% longer than adults 50+. Other factors (gender, employment, household composition/size, home type, bills responsibility and most hair characteristics) showed small or uncertain effects after adjustment. Findings demonstrate how smart metering, event-level behavioural data and robust modelling can reveal actionable levers for demand-side management, behavioural interventions and product design.