To commemorate the 25th anniversary of big data at the Capgemini consulting company, Lee Brown, head of big data analytics, and two senior consultants in the organisation’s analytics team sat down with Data IQ to talk about a quarter century of changes and the ongoing challenges clients face.
Data science is a challenge centred on business problems. It is all about finding cashable and non-cashable benefits that can be achieved by exploiting available data. These were the words of the director and head of big data analytics at consulting, technology and outsourcing company Capgemini, Lee Brown. Along with two senior consultants within the analytics team, Iain Hubert and Yigit Gungor, he spoke to Toni Sekinahon the anniversary of 25 years of data science at the organisation to discuss the changes and challenges it has seen.
Twenty-five years ago, Capgemini first created an in-house analytics consultancy after acquiring the operational research team from British Gas. The team started with 15 people and has since grown to 140. Brown's practice deals with data engineering, data science and data visualisation, and has gone through some name changes in that time. Initially, it was known as the data practice, then it became the business intelligence and data practice. After that, it was called the business information management practice, and now it is the insights and data practice.
Gungor, the youngest of the three who described himself as part of “the new breed of data scientist”, has even seen changes in the nomenclature of the industry during hsi short career. While at university, what was then called neural networks is now known as deep learning, even though the concepts and the core of it are the same. Hubert’s team went through more of a straight-forward change, starting out 25 years ago as operational research and becoming operational analytics 20 years later.
Despite these name changes, the core functions have remained the same – using data to solve problems. These problems are wide-ranging, from optimising cash-flow to bringing more efficiency to the way the water industry fixes leaks by eliminating the need for engineers to go around with trumpet-like listening devices.
However, an ongoing challenge, according to Brown, is convincing clients why they should be using data science analytics in the first place. He said it can be difficult to convey the value that it brings and, so, he often gives the example of the time he used radar data to monitor British woodlands. The client commented that he had been monitoring the woodland for 30 years and, within a fortnight, Brown was able to tell him things it had taken him three decades to learn, as well as things that he didn’t know before. Brown said that, when he says that, the penny drops and there is the realisation that data can transform the way organisations do business.
He said another challenge is deciding which technology to use - open source or cost products - as well as which data platform to use. Brown recommended using both cost and open source as well as MPP, NoSQL and graph databases so that agility can be embedded into the solutions that Capgemini delivers.
A significant challenge - and one that is relevant throughout the tech industry - is the skills gap. It is tough to find people with the right blend of maths, stats, communication and software engineering skills. The communication skills are especially pertinent because they are an enduring problem for data scientists. Sometimes, they are unable to impart their findings effectively to other business units and senior executives of the company.
Gungor said that a way in which Capgemini is solving this issue is to separate data scientists into two breeds - one being the storyteller and the other being the “actual, technical data scientist”. The storyteller is responsible for taking the outputs of the data scientist and putting them in a “nice story” that is communicated to the board and the rest of the business as efficiently as possible. This scheme was piloted three years ago and has been so successful that the client which implemented it has rolled it out from London to Los Angeles and Singapore.
Gungor suggested that it is important to improve the way data scientists communicate, as it will have a positive effect on how data analytics is utilised within an organisation. He said that the complexity of queries to data scientists increases proportionally to people’s understanding of what data science can do. “If they don’t know what you can deliver, the questions you will get will be, ‘how many?’, which isn’t extremely exciting,” he explained.
Hubert said that, even when there isn’t a designated data science storyteller, Capgemini will impart expertise that can be used after a consultation has ended. He said: “We always try to leave the clients in the situation where we’ve transferred some of our knowledge, skills and abilities to them.” When working with “slightly more back office” data scientists, Hubert has shown them visualisation techniques, as well the best way to create PowerPoint presentations that tell data stories.
Going forward, the inability of legislation to keep pace with the changes in technology and society could create another challenge for the data industry. Gungor spoke of the lack of legislation around artificial intelligence, specifically in reference to trading. He said: “There are currently a lot of algorithms trading, probably more than humans now, but if they are doing something illegal, we don’t have anything to actually prosecute.” AI is has benefits when used for decision support, as well as monotonous tasks such as converting images to text, as it is a great way to cut down costs. However, if it is used en masse it could cause a lot of issues and a lot more needs to be done around explaining and legislating it.
These futuristic issues are a far cry from the type that Hubert and Brown were dealing with at the start of their careers. Hubert joined Capgemini from British Gas where had been looking at the failure rates, stresses and strains of heavy-duty machinery. He said that, since then, the sort of work he has been doing hasn’t changed, but the tool and techniques, as well as the data he’s got to work with, have changed massively.
Hubert said that, 25 years ago, “it was all very backward-looking,” such as analysing how machines performed over five or 10 years and then drawing inferences of how one might do things in the future. He explained that, now, it’s all about being proactive and looking forward.
Brown remembered starting his data career as a consultant in 1996 at a small team in CSC where he had been focusing on data warehousing and business intelligence. There, he had done some “pretty nice things” with performance management and dashboards. Back then, the hot ticket was data warehousing and “bringing stuff together.” This meant that data could give a true enterprise view of how a company is performing. It wasn’t easy, though, as he remembered that the statistical tools were expensive and clunky while the models were “complicated things.”
His first “gig” at Capgemini in 2005 was with a performing arts society where he focused on bringing data sets together to understand how the organisation was running, the diagnostics on why things weren’t successful and what could be done to improve and generate more revenue. He also recalled that large database vendors like Oracle and Teradata were moving towards embedding statistical functionality within their extended language support back then. Brown remembered building propensity models and looking at market basket analysis for retailers and uncovering some real insights.
As time has gone by, Brown said that data has become more important for Capgemini’s customers and their focus has shifted from the collection and engineering of data to the exploitation of it. He has seen the increase in demand for business intelligence as organisations have wanted to understand the descriptive data being collected. He also thinks that the democratisation of business intelligence has taken place, which he puts down to more and more analytical tech aimed at the masses becoming available.
Brown explained that Capgemini itself has become more data-driven, having come to that conclusion by looking at the projects the organisation used to do compared to what it is doing now, as well as the challenges in the marketplace. The data team has also used internal data to improve the way Capgemini itself operates. The Paris-headquartered organisation has roughly 200,000 employees across 40 countries which led Brown to say, “we’re a people business.” They have used data to match employees to opportunities within the company and carried out a project looking at early warning signs of employees that might want to leave.
Gungor said that Capgemini has to make sure that all its clients are in the position to integrate analytics into their operating model in the most efficient way as there is not one model that fits all. “It all depends on what products you are generating, how you interact with your customers and even the culture of the company,” he said. Despite all the changes that the trio have seen, there is one thing that will not change, according to Gungor, and that is the need for organisations to embed analytics at the core of their strategy. As he said: “It is not something to improve and just do once.”