Having read physics as an undergraduate, I joined the private equity industry buying, financing and managing businesses in the retail, leisure and education sectors. We were one of the first private equity businesses to use detailed quantitative techniques to add value to our investments. But after 12 years I ended up finding PE quite annoying, so I spent two years studying microeconomics and econometrics at UCL.
Econometrics is very similar to machine learning, so I joined Betfair as principal engineer, where I learned functional programming in Scala, and implemented machine learning systems to recommend betting markets in real-time, identify fraudulent customers and predict the outcomes of tennis matches.
From there I moved to Barclays to be director of advanced analytics, where I built a team of data scientists which (we think) put the first Spark application in production in the financial services sector in London.
In my current role I am driving a transformation to evidence based decision making throughout the business. We have expertise in three areas: data engineering and automation; analytics and visualisation; and data science; and are in the middle of implementing a hub-and-spoke federated model of analytics with a billion-pound medium-term target for profit contribution.
Last year, JLR’s data and analytics team of 44 people delivered an audited profit contribution of £146 million in its first full year of operation, with two-thirds of the team in the graduate programme, and in a very tough year when the whole JLR profit was only £273 million. I was particularly proud of this because when the Chinese market collapsed and diesel engines declined, the business needed us to step up and help the transformation plan and we did. Over 250,000 people rely on the success of JLR. We couldn’t afford to let them down.
At Barclays, I was lucky enough to work for a great manager, Nick Hall. Not only did Nick show me how to build a high-performing team, and had visionary support for high-end data science, he also lived the values of the team, respect, integrity and excellence.
To stay innovative, the team does projects where we aren’t sure what the outcome will be. Last year, we built a world-leading hierarchical dynamic Bayesian model to forecast car sales to unprecedented precision and speed. We had no idea whether it would work or be adopted by the business, some of whom were skeptical. But we persisted and in a chance meeting with CEO we showed him a huge opportunity we had identified. It is now central to the business planning process. Sometimes analytics is a matter of keeping the faith, rather than expecting results.
Data and analytics are no longer “new” news, and companies are starting to demand that it produces real measurable profits, not just cool toys and “that’s interesting” insights. While there will still be teams set up around blue-sky research and technical showcasing, increasingly in 2020 data and analytics will be expected to generate returns from day one. This is a great opportunity for our field, allowing us to move from being seen and funded as a cost centre like IT to a profit centre, recognised for being a core part of product proposition.
The car industry is being transformed by the use of data to make cars autonomous, connected, electric and shared (ACES). The annual UK expenditure on transport is £200 billion, excluding the wasted resource and health impact of sitting in a car or bus for an average of 10 hours a week, and 180,000 traffic casualties per year. ACES will change the way we all think about personal transportation, motivating Jaguar Land Rover’s journey to “destination zero”, a world of zero emissions, zero accidents and zero congestion. Being smarter with vehicle data will make our societies safer and healthier, and our environments cleaner.
There is the constant challenge of getting good data on which to base decisions without embarking on time-consuming and expensive master data management programmes, which usually fail. But our main challenge is how we can industrialise insights so that the business can integrate analytical applications into a repeatable automated process. If you don’t do this, all of your analysts end up tied into supporting legacy analytics and can’t do anything new or interesting. If you manage to do it, you can turn every pound of profit contribution into maybe 10 pounds of sustainable shareholder value.