The winter of 2010 was the coldest in Britain for 300 years, causing major disruptions across the country including an unprecedented number of boiler breakdowns in people’s homes. That was despite British Gas analysing data from nine weather stations to make predictions about what weather conditions and temperatures would be.
“We went into winter 2010 thinking that we were better prepared than ever for the winter situation,” Steve Mcgivern, national forecasting, rostering and planning manager for British Gas Services told Computing at the SAS Premier Business Leadership Conference in Amsterdam last month.
“We thought we were going into winter with the perfect situation and October and November were fine. But then December arrived, the coldest ever experienced, and the demand went through the roof.”
The previous record for breakdowns in a day stood at 25,000, but during one three-day period, that peak rose from 34,000 to 39,000, and British Gas didn’t have enough staff or resources available in order to meet its 90 per cent same day service target.
In the following months, British Gas examined its infrastructure and made purchasing new forecasting systems a priority, with the end result coming in the form of a SAS forecast server.
“What we needed was a more robust forecasting system that used a lot more data, and a lot of weather information. We did a proof of concept with SAS and found they could beat the old system hands down.”
The new SAS system allows British Gas to analyse much larger amounts of data, with feeds arriving from 30 weather stations as opposed to the previous nine and temperature predictions extended from three to 10 days. Introduced last October, it allows British Gas to more accurately determine the number of staff it requires on standby in different parts of the country, depending on the weather forecasts.
“The forecast accuracy has been exceptional. We were aiming to be just above 90 per cent accuracy but pretty much every week we’ve been averaging 98 per cent, which is pretty unbelievable,” said Mcgivern.
“By having that accuracy, you’ve got the resources working when you need them, you can respond to customer needs and our same-day customer service target got back to where it should be all the way through the winter.”
Looking to the future, British Gas is keen to use another SAS tool, SAS Operations Research Software, to help it to optimise the way it allocates geographical areas to its service engineers.
“It’s probably fair to say we’ve not used a tremendous amount of science in the past in determining what a patch should be,” Mcgivern told Computing.
“But we know where the engineers live, we know where the customers live and obviously we know where the boilers are,” he said. “We involved SAS to re-optimise our geography using real science, and the early indications are that we could reduce travel by 30 per cent.”
According to Mcgivern, it will be possible for engineers to complete more jobs per day at no additional cost for British Gas.
“Travel is a big factor – obviously you want to get to a customer with a breakdown as fast as you can – but a reduction of travel by 30 per cent is giving you the capacity to do an extra job a day with each engineer, that’s 7,000 additional jobs for free,” he said.
Mcgivern also revealed that while management of staff jobs is currently done using Microsoft Excel, he’d like to move away from it, transferring to SAS software.
“Eventually, we’d like to transfer our rostering platform, which is Excel-based, to a SAS platform, then you’ve got everything in one subject area. Hopefully that’ll be in the near future,” he said.