Introducing the Chief Data Officer
Data science is a hot topic right now, with companies like DeepMind, Google and Baidu capturing the world's attention with exciting innovations that offer a glimpse into what the future may hold.
With computation and data storage becoming ever more affordable, and a huge amount of data to learn from, there's no wonder the field has matured in its identity, techniques and tools.
For senior leaders, however, the nascent and complex field (and its wealth of associated jargon) can feel completely impenetrable, leading to misunderstanding of how best to invest in it.
To grapple with these complexities, Chief Data Officers have entered executive teams. Though the role has existed before under different guises, it has grown in its level of influence, from a regulatory focus towards a remit that is broader and more strategic. Data science teams today are meant to supercharge the value businesses create and propel it ahead of the competition – and that’s a fundamental shift.
So how can CDOs make data science ‘work’? Where should they start and how can they set the team up for success? And what kind of challenges should they be anticipating?
I spoke to Piers Stobbs, Chief Data Officer at MoneySuperMarket, to get his reflections.
Making the case for data science
“I think about data science in a really simplistic way, which is about solving problems with maths and data.” With almost twenty years of experience in analytics and data, Piers is clear about the value that data can add: “We essentially make humans or machines better at decision-making.”
MoneySuperMarket has a clear mission – helping every household to make the most of its money. This north star unites people across the organisation, and Piers’s work revolves around using data towards that end. But do senior leaders understand where data science can contribute and how?
On the exec team, ‘explaining the art of the possible’ is often imperative to the CDO role, as Piers explains: “There are misconceptions, and it can happen both ways.
One is, ‘AI can solve that’, and in many cases, it can’t, the data set might not be available. On the other side, there can be assumptions that something is really hard, and that might not be the case.”
Measuring success with data
How to measure the success of data in an organisation poses a compelling challenge - the utility of data often has wide-reaching implications, and may help towards the targets of other functions like Operations or Marketing. Piers therefore believes that effective, data-driven organisations should set cross-functional KPIs.
“I think if we’re being successful, those things (data metrics) should be delivered through other business units. So for example, I’d sign up with the marketing team to improve our return on ad spend or with the product team to improve our customer conversion.”
When designed effectively, shared KPIs can incentivise key stakeholders to think and work together, and crucially ensure that data scientists are solving business goals.
“Getting that alignment with other colleagues is critical. I’m quite scathing of data scientists who create a technical thing that doesn’t solve the problem.”
Piers also warns not to underestimate processes involved in implementing solutions, and particularly the costs involved in change management.
“Often what you’re trying to build will change how certain processes work in fundamental ways. For example a chatbot might really impact how a call centre works.”
Creating teams that can deliver impact
The challenge of how best to build and utilise a team of data scientists is much-discussed at the moment.
As a nascent field, there is a sense that many organisations have invested in personnel without thinking about how they are integrated.
Reflecting on his experience at MoneySuperMarket, Piers said: “If I was building a thing from scratch, I would start with data engineers to make sure the platform works well. Then you need some good analysts to visualise the data and bring it to life, and then get some data scientists to look at some areas that might gain traction. It’s that blend of skills that you need, rather than 15 people writing deep learning models.”
The cost of siloing data scientists, removing them from customer problems or the business context, can be brutal.
“One of the biggest sources of failure in my experience is when you’ve solved the wrong problem.”
“You can have smart people going away to solve a problem and then they come back and you find out that you can’t implement it or it’s not quite the thing you needed to fix. That can be a great source of frustration.”
The future of data in business
As technology advances, companies and industries that fail to adapt make themselves vulnerable to disruption. Piers agrees that companies that don’t use data science will be left behind, but is clear that data science should be used to complement, rather than replace, the human touch.
“I’m a huge believer in human-in-the-loop – there’s a great quote that the best chess player isn’t DeepMind or the grandmaster, it’s a combination of the two, because there are still instances where that human nuance can win out.”
“So I believe that using machine learning techniques to surface a set of new possibilities and then allowing a human expert to select the most appropriate solution in a given case is a very powerful paradigm.”
Finally, what advice does Piers have for senior leaders considering a new data role?
“I would warn people against coming into a senior role without the ability to deliver... It takes some frank discussions with CIOs, and Heads of Product and Marketing about where the responsibilities are as you hire someone in, because it can be grey. So I think having clear responsibilities and a clear delivery mechanism for the CDOs is really important.”
This post was originally published on the Made By Many Blog in July 2019. Piers Stobbs is the Chief Data Officer of MoneySuperMarket.