The transition towards decentralised and decarbonised electricity systems requires a detailed understanding of residential energy demand to inform effective Demand-Side Management (DSM) strategies. Current modelling approaches predominantly rely on socio-demographic factors to capture occupant behaviour heterogeneity, potentially overlooking critical household energy usage variability essential for low-voltage grid analysis. Alternatively, we propose a methodology centred on temporal and intensity patterns of appliance usage. Specifically, we develop an activity-based energy demand model that establishes pattern-dependent links between occupants' activity schedules and appliance switch-on events. Furthermore, we evaluate our approach's ability in estimating the distribution of DSM potential within the population, comparing it with conventional approaches based on socio-demographic classes. By examining laundry and dishwashing as case studies, the results demonstrate that our approach provides a more comprehensive characterisation of energy demand heterogeneity compared to conventional socio-demographic-based approaches, offering three key advantages: (i) enabling the identification and targeting of population segments with higher flexibility potential, (ii) determining the optimal contribution of different population segments to various flexibility services, and (iii) informing the design of actionable DSM interventions. Lastly, we propose strategies for operationalizing a pattern-oriented approach in real-world contexts and emphasise the need to integrate different data sources to advance DSM research.