One swallow does not make a summer (a fact that I have found to be oh so true since relocating to Edinburgh!) nor does one pie-chart in a powerpoint make you a data driven company. When I became Chief Data Officer at DBS in 2016, I was tasked with making DBS “data driven”. How hard could it be? We had several successful transformations under our belts and surely this one would be just a rinse and repeat. As it turned out it took me three years simply to figure out what a data driven company actually was and then getting to be one was the biggest challenge we had ever faced.
Prior to 2016 we had been operating as a “powerpoint driven” company where decisions were taken in meetings by the most senior person in the room based on the content of a slide deck. So when I took on the role I asked many leaders within and outside DBS what they considered a data driven company to be. I was told that data was the new oil or water or even blood. People referenced TV shows and movies – Minority Report anyone? Every software vendor claimed that Artificial Intelligence was embedded in their products – data was clearly becoming the new snake oil. Of course as soon as we announced our intent to be data driven there were claims that we were already data masters – “We are a bank – we use numbers all the time”, “Look at the pie chart in my powerpoint”. We knew we had to do some heavy lifting in terms of addressing data quality, bringing all the bank’s data into a new tech data platform, investing massively in training and attracting the best talent. However we did not have a clear view of what the end state looked like.
Three years down the line it felt like we were in good shape – most of our data was of high quality, referenced with metadata and residing in a leading edge data platform. We had trained huge amounts of people and attracted some great talent. Because we adopted our tried and tested approach of encouraging people to just dive in and have a go, we had over 100 projects underway and were starting to deliver real benefits. However what we really had was just a bunch of projects. We were not running the company day to day using data. Powerpoints still ruled the roost.
Then the epiphany came. We were inspired by the evolution of data use in Formula 1 Motor Racing. Over the past few decades the sport had gone from using signboards at the side of the track to sophisticated instrumentation on the cars that transmit huge amounts of data during the race to allow the pit-wall crew to make real time adjustments to strategy. In between races experimentation and data analysis’s results in continuous incremental improvement. We asked ourselves what would it take to run our business this way.
We realised that to be data driven we needed to completely re-imagine how to run the company. Despite the progress we had made, this final step was going to be the largest and toughest. We needed a new approach so we developed the following 5 steps:
- Be ultra specific about what constitutes success of the business. What is the exact outcome measure that needs to optimised. In F1 it is obvious: “Did you win the race?”. In business it can be less clear. Through discussions on outcome measures we highlighted some mis-alignments in our strategies that we were able to iron out.
- Identify 3-5 drivers that have the biggest impact on the outcome. I am no expert but I would imagine winning a F1 race is impacted by engine performance, tyre strategy, driver capability, pit stop timings etc. In business there can be a tendency to analyse the outcome – in review meetings leaders tend to drill down on revenues by geography or product rather than focus on what they can influence i.e. the drivers. Furthermore there is scant thinking about the relationship between the drivers and the outcome. If training improves the performance of a sales team, what kind of training works best and by how much? A data driven company seeks to continually improve their understanding of the causal relationships between drivers and outcomes.
- Identify opportunities to apply machine learning to improve the drivers’ impact on outcome. Machine learning can beat humans in predicting outcomes given the right data. Therefore developing models that take some of the guesswork out of people’s job makes sense. Who is the best customer to call next? Does this transaction look suspicious?
- Relentlessly experiment to test hypotheses. In order to continually optimise, previously held beliefs need to be challenged. A data driven company is an experiment machine and focuses on accelerating the speed of learning by investing in experimentation infrastructure, processes and training. Which leads to the big one….
- Change the leadership culture to start asking questions rather than giving answers. The big tech companies who excelled in an experimentation-led approach consistently told us that only 1 in 20 hypotheses are proven to be correct and therefore leaders should expect to be wrong. A HIPPO culture where the HIghest Paid Person’s Opinion always wins results in learning opportunities being blocked. We therefore encouraged leaders to ask 2 questions of their teams. “What experiments are you running next?” and “What did you learn from the last set of experiments?”. Oh and this helps create a culture of psychological safety – an essential ingredient for innovation.
- Spend an inordinate amount of time aligning leadership understanding of what is meant by “being data driven”.
- Focus on improving drivers and not acting directly on outcomes.
- Expect to be wrong and therefore focus on test hypotheses through experimentation.
- Becoming data driven is a culture shift more than an investment in technology, process and data scientists.