Round table: Big data, small costs

This is the age of information. Our society is collecting more data, more quickly, and from more of sources than ever before. The variety and complexity of that information can be confounding, but there are also increasingly-sophisticated tools available to analyse and process data. That technology creates the opportunity for organisations to tap into a reservoir of evidence that can aid decision-making and lead to better outcomes.

As data is getting bigger, government is getting smaller. The pressure to deliver improved services with declining resources makes it more important than ever that government finds new ways of utilising its data to the full. In partnership with data analytics provider SAS, Civil Service World has been bringing together groups of civil service information experts to discuss various aspects of ‘Big Data’ – and this round table focused on how it can be used to make government smarter and more efficient.

Utilising data to improve efficiency
Several of the assembled civil servants represented departments with a substantial wealth of data to draw upon. For example, Siobhan Carey, chief statistician at BIS, can access a database listing every person who’s undertaken further and higher education in the last 20 years. And Chris Kershaw and his colleagues at the Home Office gather data on more than 100m people who cross the UK’s borders each year.

The data that government can utilise falls into two categories: operational and transactional data held by government; and information from non-governmental sources, such as social media, which can be analysed to provide government with useful insights. The first category provides the most obvious opportunities for improving efficiency within government, and a number of attendees cited examples of programmes where data analysis is being used to improve their operations.

“From DWP’s perspective, one of the main advantages of using this data is to prevent benefit fraud,” said Emily Gilmore, the head of the department’s legal team. “In the same way as HMRC are pursuing ways of looking at a huge variety of data to tackle tax fraud, we can look at the same types of data to tackle benefit fraud.”

Her DWP colleague Corinne Richardson added that the speed at which information can be analysed is crucial: “To stop fraud, you need to be on to the data at its freshest before the person moves on to something else,” she commented.

“Anything that makes us better improves efficiency. It is not just a narrow band of money-saving.” Danielle Mason, Cabinet Office

Barry Proudfoot from the FCO said that data analysis helps his department to allocate resources: “For example, we collect information on [levels of activity in] our consular network to work out whether we need to put more officials in Spain in the summer time and Thailand in the winter time.”

MoJ has derived similar benefits from information analysis, added statistician Lucy Cuppleditch: “You can use it for planning because you can follow the flows through your system. At the moment we are taking recorded crime data from the Home Office so that we can track the flows going through the court system and into the prisons, and understand better when we are going to have a peak in demand,” she said.
Chris Kershaw from the Home Office suggested that examining data on casework might help departments to save money: “If we look at the patterns of how we deal with cases, it might help us to understand how to process them more quickly. There may be cases that can be dealt with without expending so much of our resources,” he said. He added that the Home Office is working jointly with DWP and HMRC to analyse information held by the three departments to identify cases of sham marriage.

Amanda Gardiner from SAS said her company has been working with the Home Office to investigate how border control resources can be allocated more efficiently: “You may not need six people manning desks for a flight coming in at a particular time, because 90% of the people on that flight will have the right visa,” she said.

Profiling for efficiency
That line of discussion led to a debate on the usefulness and desirability of government using the personal data it has about people to determine how much time and resources to expend on scrutinising them – a technique known as ‘profiling.’

Simon Dennis, client director for central government at SAS said that resistance within Whitehall to the use of profiling is weakening: “When we were working with the Inland Revenue, there was a view that you could create a statistical model that would tell you whether someone is more or less likely to be [non-compliant with tax law],” he recalled. “A lot of the civil servants at the operational end said: ‘We shouldn’t do that. We don’t know that person is bad so we shouldn’t treat them any differently.’ It was when the cuts started to bite that people said: ‘We have to do this. We have to use every tool at our disposal to reduce our chance of wasting time’.”

DWP’s Richardson warned that successful profiling relies on good quality data: “It is about what the evidence base is for the profile,” she said. Meanwhile Mattthew Dovey from Jisc Solutions, which provides IT infrastructure for further and higher education, suggested that profiling can still be controversial in some circumstances: “The classic example that hits the headlines is that certain elements in society feel they are picked on more in security checks,” he said. “It is more hidden if you are looking at someone’s tax return. I don’t know if my tax return is being more scrutinised than others, but stop and search is much more visible,” he said.

Using data from outside government
Other less obvious benefits to efficiency may accrue from the use of data harvested from sources outside government. As Danielle Mason, head of the social research profession, said: “Anything that makes us better improves efficiency. It is not just a narrow band of money-saving. We are talking about our operational data and how we can do more things with it, but that is still fairly traditional. There is another side to big data that is a bit less well understood in government, which is about data that doesn’t belong to us and the opportunities it presents in terms of policymaking and horizon-scanning.”

“You have to find people who know about the topic to look at a sample of data to see if any errors have been made.” Savania Chinamaringa, Defra

Navroza Ladha, a DWP lawyer, cited a very recent instance of the use of non-governmental data: “Before I came out, one of the headlines on the BBC news website was: ‘Ebola: can big data analytics help prevent its spread?’ It was saying that mobile phones are proving to be a rich source of data. A Swedish non-profit organisation was able to draw up detailed maps of typical population movements in Senegal, so they knew where to set up the treatment centres. That is an example of thinking outside the box and looking at not just our data but also other people’s data,” she said.

Kershaw raised doubts about the reliability of information mined from social media, however: “Because of the way it is collected, it is not necessarily representative and it could mislead you,” he said.

Data quality
One of the big challenges facing data analysts is information’s reliability. Carey from BIS said: “The main issue in government is the complexity of the data. We have a lot of data that has been collected over a long period of time and it is in different formats.”

Dominic Rowley, head of governance for the civil service pension scheme, pointed out that recent reforms are placing a heavy burden on government data: “For the pension schemes, the data is largely historic. For 40 years, as long as the last three years were alright it didn’t matter what had happened before then. We now have to get things right on an ongoing basis, and we are finding that the level of data that is held by employers is way below what we would like.”

The attendees agreed that how ‘clean’ data needs to be depends on its use. “The question you have to ask is: does the lack of data quality make it fit for doing only a limited analysis of trends, or is it 100% accurate and fit for purpose for getting down to the nitty-gritty?”, said Richardson.

Determining information’s reliability is not straightforward, however. “The way we do it at Defra is to look at system errors,” said the department’s data lead Savania Chinamaringa. “There are systematic things like missing entries that you can use, but misinterpretation is difficult to predict. You have to find people who know about the topic to look at a sample of data to see if any errors have been made; it is incredibly difficult to know what the quality of data is.”

Technology and skills
Does government have the capability to make the best use of its data? A number of attendees stressed that neither policy expertise nor analytical acumen alone is sufficient. “This requires multi-disciplinary teams. You’re not going to get all of the skills around the analytics and the interpretation and the topic knowledge in the same person, so you need to bring teams together,” said Carey.

She expressed confidence that new graduates recruits have data analysis skills. However, at present they lack the required technology, she said: “They come into government and don’t go anywhere near that stuff because we don’t have the tools for them, but there is certainly a latent skill set there.” Both Carey and Richardson said their departments are in the process of building an outsourced data analysis capacity.

“We look at data in very siloed ways in each individual department.” Chris Kershaw, Home Office

Julia Jagelman from DWP agreed that technology rather than skills is the principal barrier to progress in her department: “I would like to see the analytical community in our department having a look at what they can do with big data, but it comes back to the capacity. I think if we had [the IT] built, they would be interested,” she said.

Dennis admitted that the IT industry shares some of the responsibility for slow progress in introducing data analytics into government: “We can’t just blame statisticians for being timid,” he said. “IT failures have been a joint effort between the IT industry and government and it has made people afraid of doing these projects – but now it is cheap. What you needed a mainframe to do three years ago, you can now do on something that costs £5,000 in half the time. That is how much things have changed.”

Changing the culture
When the attendees were asked whether cultural factors constrain the use of data analysis, Chinmaringa heartily agreed: “Big data is at the cutting edge. We are experimenting. If we try it and it doesn’t work, we try it again,” he said. “Culturally, that is not what government does. There’s always a fear about the reputation of the government and the department. There’s a feeling we would rather be safe than sorry.”

In two recent cases, predictions based on data analysis have gone awry for Defra. It underestimated the risk of flooding this year, and badger culling proved less effective than anticipated. “People are nervous about that kind of exposure,” said Chinmaringa. “In the private sector, you can make a mistake and the company can collapse – but governments cannot collapse. People expect us to be careful and rigorous.”
Kershaw observed that a change of approach is also needed to encourage data sharing: “We look at data in very siloed ways in each department. There is no cross-government view on data as a whole, and that is where there could be some value in joining up, although you have the legal and data sharing headache that goes with that.”

Gilmore provided a lawyer’s perspective on the legal barriers to making greater use of information: “A lot of what we understand big data to be about is, on the face of it, at odds with data protection,” she said. “There are issues with accuracy, proportionality and re-purposing the data, so it is important to involve lawyers at an early stage – but those things aren’t show-stoppers. There are ways around all of them,” she advised.

Several attendees called for more examples of best practice in data analysis to be publicised. Carey suggested that the Cabinet Office data science programme is making progress in that area. “You will not persuade anyone by telling them what you can do, but if you show them what you have done and how that can be used then you’ll win their hearts and minds,” she said.

Government has yet to do more than scratch the surface of the opportunities presented by data analysis. If improvements in technology and changes in civil service culture develop alongside each other, it may yet make a big contribution to more efficient administration.