Accountants and data scientists - collaborating for success

There is no standard definition of data science or established curriculum. However, a useful working definition is ‘Data science involves extracting, cleaning, organising, combining, analysing and presenting data to help organisations learn, make decisions and manage risk. It involves the use of scientific, mathematical, statistical and machine learning techniques. Tools used range from off-the-shelf software packages to programming languages to build bespoke models.'

A data scientist has been defined by Drew Conway, CEO and founder of Alluvium, as someone who combines maths, statistics, hacking/coding skills and domain knowledge (e.g., business knowledge).

Centrica internal audit now employs data scientists within their team and we are extremely grateful to them for sharing the knowledge they have gained.

• Data science is a fast developing, young discipline. It has the potential to strengthen business decision-making, financial management, risk management and internal audit. However, because the field of data science is so new, we are still learning the best way to realise this potential.
• The degree to which accountants will engage with data science will depend on their  knowledge, skills and attitudes. Attitudes are perhaps most important. If accountants have the passion and curiosity to use data in their work, they will find ways to develop the knowledge and skills. Further detail on knowledge, skills and attitudes is provided in the next section.
• To achieve a step-change in how finance and internal audit departments think about and use data, they will probably need to bring data scientists into their teams. They will come from a range of backgrounds eg, Centrica internal audit has a team member who is qualified in advanced computational chemistry. Rebadging accountants will not be sufficient and centralised data science teams may not have time to support finance.
• As accountants start to interact regularly with data scientists and work jointly on projects, they will learn new skills, as will the data scientists. This will result in a blurring of roles over the next five to 10 years.
• Accountants can play a key role in providing the bridge between data science and the business because of their commercial skills and broad training. However, it should be noted that data scientists are also aiming to develop more rounded business skills.
• At the moment, data science is unregulated. There is no established professional body for data scientists and career paths vary widely. This makes recruiting appropriate data scientists challenging.
• The way in which data is used for decision-making, risk management and compliance will vary depending on organisational circumstances and culture. We noted more organisations are appointing Chief Data Officers (CDOs) who will have a key role to play in shaping the approach to data and getting the right skills in place. However, there may be a shortage of people with the necessary experience and skills to take on the CDO role.
• In many organisations, the use of data is still in its infancy so there are many benefits to be gained by relatively simple uses of data. For example, accessing and interrogating data at a more detailed level, connecting disparate sources of data and slicing and dicing data in different ways to look for patterns and outliers.

• Accountants with high levels of curiosity and passion around how data can help solve business problems will be best placed to make the most of data science. They will also need to be agile, imaginative and willing to move outside of their comfort zones and experiment with new ways of doing things. For example, while developing hypotheses around business problems and looking for the 'right answer' will remain important, accountants will also need to 'play' with data and see if something unexpected and useful emerges.
• The depth of knowledge that accountants will need to know about data science will vary depending on organisational needs, resources and structures. For example, in large organisations with significant numbers of highly skilled data scientists, accountants may only need to know enough to collaborate effectively with data science teams. The data scientists would be trusted to execute projects properly. However, given the shortage of data scientists, the lack of a defined data science profession and the increasingly user friendliness of data science tools, accountants could expand their knowledge and skills to undertake data science tasks. It could also be argued that as organisations shift towards more sophisticated uses of data, accountants will have to broaden their skills in order to maintain business partnering roles. If not there is a risk accountants will be forced back to narrow specialist roles.

We identified the following data science skills and knowledge, ordered in increasing levels of detail:

• Opportunities and limitations of what can be achieved through data and how data science can add value. Understanding and defining the business problems data can help to solve.
• Interpreting the outputs produced by data analytics. This includes understanding data provenance, modelling assumptions, inherent biases in the analysis and, perhaps most importantly, what decisions can justifiably be made based on the analysis.
• Presenting and communicating the results to the business, including the use of visualisation.
• Awareness of the data landscape, different data types, what data might be useful and where and how it can be obtained.
• The importance of data security and the relevant organisational policies and processes.
• Knowledge of the wide range of statistical and data analysis techniques available and their uses, strengths and weaknesses. We did not discuss whether this would extend to sophisticated techniques such as k-means clustering, random forests, neural networks etc.
• How to connect and clean up data. NB this can take up to 80% of the time for a data science project.

• In order to obtain further investment in data science capabilities, finance and internal audit departments will need to become more systematic and effective in communicating the value being achieved from their use of data science. This requires marketing and PR skills.
• Internal audit teams, while drawing on data science to improve their work in general, will also need the skills and knowledge to provide independent assurance on the use of data, analytics and algorithms elsewhere in the business.
• The ethical issues around data use and algorithmic decision-making are becoming increasingly important. Accountants will need to further develop their ability to identify and resolve ethical issues around privacy, bias and discrimination.
• As mentioned above the main way accountants will learn about data science is by working with data scientists and using data science in their work. Centrica helps accelerate the use of data science by creating time for team members to experiment with data to solve business problems. The data science team within internal audit also sends a weekly email to promote curiosity in exploring how data can be used in everyday life, such as minimising commute times.

The ICAEW Library & Information Service provides full text access to leading business, finance and management journals. Further reading on developing your data science capabilities is available through the resources below.