Data leads water utilities in customer behaviour

Posted 7 July 2016

Dr Don Perugini, Executive Director, EYC3, Asia Pacific Advisory, Analytics & Enterprise Intelligence
Dr Michelle Perugini, Executive Director, EYC3, Asia Pacific Advisory, Analytics & Enterprise Intelligence

water faucetWater utilities are facing disruption and greater risks due to changing consumer behaviour and demand, and can no longer rely solely on engineering or on building infrastructure with the assumption that their customers will purchase as they always have.

Rising consumerism and developing technology is resulting in customers expecting better service, with options to seek alternatives (for example, grey or rain water storage, efficient appliances, landscape changes or recycled water) to better manage and control their own water demand.

Utilities need to become more customer-centric, using data and analytic tools to better understand disruptive impacts, and manage their assets and risks more effectively. 

Better demand modelling approaches will also empower utilities to address numerous challenges, from ageing infrastructure to tighter environmental regulations and increased pricing pressures.

Signs of future disruption in the water industry

Demand for water, both globally and in Australia, is declining due to a range of factors, including use of more efficient appliances, higher density housing and behavioural change, from changing climate conditions and the environmentally conscious consumer.

Drought is becoming a more common occurrence due to changes to climatic conditions. A few years ago Australia emerged from a drought and currently California is experiencing one of their worst droughts in history.

Droughts have two impacts for utilities and consumers. Firstly, water shortages require consumers to change their behaviour to temporarily conserve water throughout the drought period. Identifying and quantifying the conservation measures that are most effective across a diverse population of consumers can be challenging without the right data and tools.

Water conservation, such as water use restrictions, results in rapidly changing demand. This can be difficult to forecast with traditional models that rely on past demand or behaviour, resulting in greater uncertainty for utilities during restriction periods. Forecasting errors of up to 30% have been experienced by utilities in the past.

Secondly, temporary conservation measures can have long-term effect on demand, as behavioural change during drought can result in persistent, long-term changes to behaviour. 

Additionally, easing water restrictions can create further uncertainty around bounce-back in demand. Accurately forecasting bounce-back without tools to explicitly model the broad range of consumer behavioural factors can be challenging.

Even though the decrease in demand for water is perceived as good for the environment, it is vital that water utilities understand why demand is changing and what the future trend will be in order to support operational and strategic decisions. In this context, the development of demand modelling and forecasting tools is particularly important.

Traditional forecasting approaches are limited in predicting these trends, while building infrastructure based on today’s demand can result in over-investment and financial risk, ultimately increasing bills for customers and driving down demand even further.

Current approaches to demand modelling and data intelligence

Water utilities use a range of demand forecasting approaches, including the constant rate model, statistical and econometric models, and neural networks.

The constant rate model is one of the simplest models to forecast water demand. The model uses a constant rate of water consumption per sector (such as capita, industry) unit. This engineering approach is used to obtain a first estimation of long-term water demand in a sector according to demographic change and economic development.

This model is well-suited to areas where water demand is governed by strong demographic and industrial changes, but doesn’t enable the evaluation of policies such as water price changes, water restrictions or modifications of water demand due to behavioural change.

Other forecasting approaches rely on using past behaviour and demand data to forecast future behaviour and demand. However, in a rapidly changing or disrupted consumer market, the future is likely to differ to that observed in the past, meaning errors in forecasting can result.

Additionally, these models rely on a limited set of variables that influence demand such as weather, rates or price, restrictions and drought, population, economy and appliance efficiency and code savings. Each variable represents an aggregation of consumer factors and how they interact, over-simplifying how humans behave.

These forecasting approaches cannot accurately and comprehensively model consumer behaviour. Furthermore, this limits the types of scenarios and interventions that can be tested using these models, such as testing the impact of specific conservation measures like changes to garden watering frequency and times.

New customer demand modelling and data intelligence approach

Simulait is a new method of forecasting demand based on behavioural microsimulation, which comprehensively models consumer behaviour and accurately forecasts demand, particularly in the event of disruption such as drought. The model can also test a broad range of scenarios and interventions to assist with predicting and influencing consumer behaviour.

Developed by EY, the water microsimulation demand model has been applied globally, from assisting with water conservation and forecasting in California, to assisting with capital planning and regulatory submissions in Australia. As model variables are already configured using a wide variety of public data sources, minimal data is required, allowing users to configure as few or as many variables as needed.

The behavioural microsimulation approach comprises a detailed bottom-up, appliance-based model to simulate individual consumers, the appliances they have and their usage behaviours. This includes usage behaviours inside and outside the home, and changes in behaviour in response to given external influences such as weather, restrictions, conservation communications and new rates.

The model incorporates a range of publicly available data sources, including census data and population dynamics, to build a profile of individual households. Market research and end-use studies are incorporated to define behavioural rules for household activities and, ultimately, appliance usage. Uniquely, it allows water demand to be derived independent of historical demand and scenarios can be tested to quantify the effect of new disruptors or policy interventions on consumers.

Lessons from overseas: France and California

The microsimulation has been used to analyse consumer demand in Australia, France and California. In France, we observed the model accurately reproduced annual water consumption from 2000 up to 2011, including tracking the declining trend in demand over the 11 years. As the model was only calibrated for the first year (2000) and accurately forecast the remaining 10 years, this demonstrates that microsimulation is ideal for long-term demand forecasting.

In California, the model was able to track rapidly changing demand as a result of drought and water-use restrictions. A range of scenarios were tested, including effectiveness of rebate programs, conservation communication to the community, impact of catastrophic events (e.g. earthquakes) or new regulation, and bounce-back in demand given a hypothetical scenario of the drought ending within a specific time period.