Southern Water is transforming its approach to data management in order to improve decision-making, and this has involved pooling its data workers into a centralised team.
Amy Balmain, head of data exploitation at the water company, joined the organisation from the Pensions Regulator in late 2017, and has also worked in IT at John Lewis. She arrived at Southern Water with a mandate from the then chief data officer, Peter Jackson, to bring the organisation’s data specialists into one team in order to get efficiencies of scale.
Previously, Southern Water had what Balmain described as “pockets of people doing data and analytics work, but within individual departments”. And they had a deal of “technical debt”, partly in the form of essential business processes and reporting processes being in spreadsheets and Access databases, generating data whose quality was hard to assure. This problem, if ignored, would have stored up costs and difficulties for the future.
But it is not just a question of better tooling, says Balmain, important though that is. Southern’s team have found joy in the use of cloud business intelligence (BI) technology from Birst and extract, transform and load (ETL) tooling from Wherescape.
The former has provided geo-spatial business intelligence “out of the box”, and delivered £250,000 in cost savings identified on day one. The latter has delivered speed and ease of use that has impressed the team. “We reckon it’s about 20 times faster than older environments of building code, in the likes of Ab Initio or Informatica,” says Balmain.
The company has also turned to Microsoft Azure for the cloud storage of data science activities.
But the main drive has been to get a better-skilled workforce with a greater understanding of data in the team, and to improve data literacy more widely in the organisation.
“One of the advantages of centralising my data team is that we’ve already got proven managers who are able to deliver for the business”
Amy Balmain, Southern Water
Why this emphasis? To gain a better understanding of how the organisation’s people and assets are performing, says Balmain. An example would be assessing pump efficiency, when many pumps are named differently.
Balmain presented on the work she has led at the IRM UK Enterprise Data, Business Intelligence & Analytics conference in London late last year.
She described how Southern Water manages the supply of water and waste management to 4.5 million people in a geographic area that covers Hampshire, the Isle of Wight, West Sussex, East Sussex and Kent. Its assets include 83 water treatment works, almost 14,000km of water mains, 2,375 pumping stations and 39, 6000km of sewers.
The company’s data is varied and complex, some of it about assets that are 200 years old. It has 80,000 telemetry points returning data every 15 minutes. And its thousands of call centre staff deal with thousands of customer calls, generating audio data that can be analysed to improve customer service.
Balmain, who reports into the IT function at Southern Water, manages a data team of about 25 people that works both on its own and with IT to improve decision-making throughout the organisation, which employs nearly 3,000 people and has about 1,500 contractors in its ambit.
Speaking of the water industry more generally, Balmain makes the point that 20% of energy in Europe is “used to move water around”. She adds: “If you cut that down to 19% or even 19.5%, you could probably shut down a few coal-powered energy plants.”
And reflecting on the relationship between IT and the business, in the business world more broadly, outside of the specific context of Southern Water, and where data is concerned, Balmain says this is shifting, and will continue to so do.
“I’d say IT, as a function, has historically dictated to the business what it thinks the business needs in terms of IT, and that has been fine because for a long time, IT were the only people who understood how business systems worked,” she says.
“But if you move to a cloud environment and you’ve got an HR director who comes in who has done several HR implementations and knows who the systems integrators are who are good at this, and knows what the business requirements are, then what becomes the role of IT? Is it just providing end-user equipment, or is it as providing and maintaining a governance structure?
“And how does data management contribute to that? We’re going to be looking at massive amounts of change in terms of where the various responsibilities lie. That means a bit more upskilling of people who will be able to make decisions – whether it’s financial systems investment decisions or data quality and governance matters.
“For me, one of the advantages of centralising my data team is that we’ve already got proven managers who are able to deliver for the business. It’s building that business relationship [with them] – that’s the key thing. If you are seen as other or outside, then people are not going to want to use you – they will to use their own services.”
Balmain says her data team at Southern Water has a couple of “pure data scientists” and a “layer of people who are very good at doing deep insight, understanding the principles of statistical modelling, and who understand the restrictions in the data and are able to move quite quickly to deliver unrepeatable pieces of deep analysis”. R is used extensively in the team. she adds.
However, Balmain is sceptical about data visualisation tools such as Tableau. “There are only so many visualisations that make sense to people before you start really going into infographics,” she says. “Most people still want to see something either in a geo-spatial arrangement or in a set of standard visualisations that they can understand. I think the weirder and wackier that things start to get, the less people understand.
“For me, Tableau is not an enterprise tool, and I’m interested in working for the enterprise, not necessarily for a few people who can do something whizzy with Tableau and then not necessarily be able to move it into a production-ready state.”
Balmain adds: “We can build all the stuff that we want, we can produce endless reports, but if the people receiving the information don’t know why they’re getting it, or don’t understand the variability within the data or which bits matter and which bits don’t, then it’s pointless.”