UKWIR
IMPROVING THE UNDERSTANDING OF HOUSEHOLD WATER CONSUMPTION
CLIENT
REQUIREMENT
UKWIR wanted to understand household water consumption better, leading to insights to help address one of its Big Questions: “How do we halve freshwater abstractions in a sustainable way by 2050?”
CHALLENGE
Dealing with large quantities of data of highly varied types, granularity and quality before drawing out meaningful and useful insights.
SOLUTION
Using data driven analytics and machine learning techniques we developed the framework that helps us understand household consumption of water by building customer profiles of demand.
Understand your consumers to create an impact on them
“How do we halve freshwater abstractions in a sustainable way by 2050?” This is one very ambitious and tough question that the UK Water Industry Research (UKWIR) set out to answer in its Big Question programme.
To begin to answer this we first need an improved understanding of water consumption by people and businesses. This can then help with the management and planning of our precious water resources as well as developing a means of targeting water efficiency measures.
Data. Data. And more (messy) data
How do you build a picture about people’s water use? What if it is the whole population you need to consider? How do you differentiate between different people? The answer is to analyse data to extract insights to understand different behaviour types. The more data, the clearer and more relevant it is, the more accurate picture you’ll be able to build.
This was the case in this highly challenging project. We were provided data from several water companies, data that spanned various levels of granularity, from annual billings to 15-min smart meter data. There was a big variation in data quality, with many duplications, missing values and errors. But to obtain solid insights you need data that is clean and well-structured and that makes sense – this provides the solid foundation for the analysis.
The trigger point learnings from the customer segmentation
This was a collaborative project with three companies working together. Mease consultants worked with water companies to identify data sources to use and advise us on the specific meaning of the information. Decision Lab carried out the data analysis and created algorithms to clean, process and extract insights into consumer behaviour. To this end, consumer segments were identified, as well as specific features of each group were explored and their associations with patterns of demand were determined. Project leads, HR Wallingford, then employed the research outputs from Decision Lab’s data analysis to develop an integrated representation of current and future water consumption and what drives it.
Applying Machine Learning to see the People’s behavioural patterns
After carrying out a thorough clean-up process on the raw data, we were able to start building a clear picture out of the vast and disperse inputs of information. We relied on a data driven approach that used data science techniques to extract insights from the large amounts of information from the water companies. This identified key features in the data for us to focus on. We used machine learning techniques to cluster the consumers into groups of distinct water usage profiles. The clusters that emerged as well as their characteristics were explored, and we identified a visible pattern of households moving between clusters, for example between different seasons, months or days of the week. This finding suggested that water use behaviour patterns are more dynamic than previously anticipated. Furthermore, we found that the highest water using groups make up 10% of households but consume 25% of total water – an insight that is valuable for targeting those consumers with water efficiency campaigns. Overall, we developed a powerful approach for understanding consumer demand through identifying groups of distinct water consumption profiles and exploring the drivers of such behaviour.
A lively spring of insights for future gains
As Paul Merchant, UKWIR Programme Lead and Supply Demand Manager at South West Water, says: “This project has brought together a huge amount of data on household water use and gives water companies a tantalising glimpse of a completely new way of understanding and reducing demand for water. This project supports the idea that the effect of customer behaviour has a bigger impact on water consumption than socioeconomics or weather. This provides potential new avenues for how companies target their water efficiency activities and assess the likely impacts.”
Let’s collaborate
Send us an email, to discuss a new project.
We’re a team of innovators who are excited about unique ideas and help companies to create amazing solutions.