We held our first ‘meet the data scientist’ event last night, where we brought together members from the IT and Finance and Management faculties with data scientists. The aim was simply to facilitate conversations, and thereby bring to life what data science is, different ways of thinking about data and the implications for accountants.
The discussions were very varied, and I’ll just capture here a few of the most interesting points that I picked up through the evening. What is data science? The term certainly covers a diverse range of activities and outputs, but in essence, data science is the intersection of technology, statistics and business, or domain, knowledge. The opening presentation from Jérôme Basdevant from eRevalue had a great Venn diagram that showed how these three disciplines interact.
The key difference with traditional data analysis, for example, is the addition of more technology and data capabilities. This enables, amongst other things, the interrogation of much larger amounts of data, analysis of data from different sources and more intelligent data samples. The statistical dimension enables organisations to apply research techniques in artificial intelligence, such as machine learning and natural language processing, to real business problems. Different ways of thinking about data We talked a lot about the different ways that data scientists approach problems. For example, there is much greater emphasis on a scientific approach based on proving or disproving specific hypotheses. But this raises new questions in a business context of what ‘proof’ means – in the context of the investigation of fraud by internal audit, for example, what is the threshold for proof and how do data scientists manage the risks of false positives and negatives? Participants also raised the importance of dialogue, iteration and being open in questioning. It is not always possible to answer the question, for example, if the required data does not exist. Furthermore, the risks of groupthink and confirmation bias mean that it is important not just to look for proof that you are right in your views or ask questions in order to get a specific answer. What are the skills implications for accountants? We talked about three groups - the suits (business people), geeks (the tech people) and nerds (the stats people). In practice, no one can combine expertise in all of these things and therefore building meaningful dialogue is important. That needs some knowledge of stats principles and techniques, for example, although not enough to actually ‘do’ the stats work. It also needs good communication skills – listening and empathy for example – as well as the confidence to ask questions to technical specialists. Overall, it was a fascinating set of discussions that helped members to get a better handle on what data science is and what it means for them. We will be reflecting on all the feedback from the evening to inform future activities in this space.
It was certainly a very interesting and wide ranging discussion.
An open question for me is how much you need to " 'do' the stats” in order to understand them and interpret them?
In part we develop our financial instincts from the early training and experience we get in the detailed construction of financial reports, management accounts etc. We gain an in-depth knowledge of the assumptions and estimates involved in accounting analysis and the resulting limitations. Non-accountants trust us to consider these limitations when advising on decisions. We could of course do the same and trust others when it comes to using statistics/sophisticated analytics. Alternatively we could aim to develop a more hands on understanding. This may be particularly important for accountants at an early stage of their career.
What’s your view?