Prevention is the best cure, but with 150,000 emergency calls to respond to each year the London Fire Brigade does not have the time or resources to perform safety checks on all of the capital’s 3.3 million households.
It therefore needs to prioritise areas where fires are especially likely or would be particularly damaging. “We need to go into households where we know it’s higher risk,” explains Riana Smit, senior fire analyst, “otherwise you will just waste your time.”
It made sense to identify these high-risk areas using statistical analysis, and at one point the organisation used the R statistical programming environment to develop its own analytical applications. But that project fell by the wayside when the principal analyst left. “We couldn’t really find people who were used to programming in R to be able to maintain it,” recalls Smit.
In 2008, though, the impetus to analyse at-risk locations was renewed by the organisation’s London Safety Plan, a three-year initiative to reduce the incidence of fire accidents. “The plan stated that we would do some prevention activities specifically for dwelling fires,” she explains. “We couldn’t do our prevention targeting effectively without having some information – data that drives where we’re supposed to go to.”
This time, the London Fire Brigade sought tools that would be more accessible and flexible than its prior in-house efforts; in the end it selected software from SAS.
The brigade’s analysts use the tools to integrate external data, such as geographical information and demographic data on income and deprivation, with its own risk assessments of the likelihood of fire incidents and the integrity of buildings. “We don’t try to predict the number of fires someone will have next year,” Smit says. “What we try to do is use the model to distinguish between areas and identify areas that are higher risk than others.”
The resulting picture of at-risk neighbourhoods informs which households will be checked for working smoke alarms and fire exits, and where awareness advice will be provided.
London is a city that is constantly changing, with buildings erected and demolished on a daily basis, and the data can become outdated quickly, Smit acknowledges. “We know there are going to be pockets of anomalies.” The analysis therefore takes a long, multi-year view of the data, which Smit says “smoothes” these anomalies out. “If you had ten fires in one year in a hospital, and then it’s closed down for some reason or another, over a three-year period that [incident] count will be low. It won’t be zero – but it won’t have that high-risk profile anymore.
“Behaviours change over a long period – they’re not going to change over just a year,” she adds.
It is hard to quantify the precise outcome of its analytics project, says Smit. But she adds that it certainly broadened its prevention activities – the most powerful tool it has in fostering greater risk awareness. “We’re doing the best we can to help our staff in the best way to reach the right people to do our prevention activities.”