Data science is top of the digitalization agenda for internal auditor functions, with quantitative models and algorithms in the spotlight. But I can clearly tell, from my own experience practicing data science in internal audit, that knowing the math isn’t enough. Among the lessons learned and many “scars” along the way, there are three key takeaways for auditors and data scientists interested to make the best use of data science in internal audit.
Never Take the Problem As-is
Analytics start with understanding what problem must be solved and everything around that problem. But with internal audit, most of the problems to be solved depend upon knowledge of business process, data and circumstances that auditors are not involved with on a day-to-day basis. Usually, auditors use their partial knowledge and conception to frame the problem, and then pass it to data scientists who unknowingly accept and try to solve the problem with all the “mighty” techniques but never get the answers desired.
Tips for data scientists: Before any analytics endeavor, engage with people from areas of audit interest to understand how they are handling the problem now and what metrics to gauge their success/failures. Make sure the problem given is really the problem that needs solving.
Get a Seat at the Table
Too often, data scientists are approached with a project only after the project is scoped. But analytics is not just about crunching numbers; it is also about identifying “sweet spots” fit for the techniques and data available. Auditors may have lots of problems to be solved, but they are not fully aware of what’s possible to solve these problems or are distant from new techniques that are most likely employed in a business rather than audit setting. Consequently, the best analytics opportunity is missed in project planning. The more embedded and collaborative a data scientist is involved in project planning, the more effective he/she is afterwards.
Besides, becoming a truly data-driven auditor requires a fundamental change to legacy systems, processes and outdated mindsets in delivering audit projects. For data scientists, coming up with innovative analytics ideas is the easy part; implementing these ideas and integrating them with the existing audit approach is the true test of a data scientist in the audit realm.
Tips for data scientists: Never mind to get your hands dirty – proactively talk to auditors and request a seat at the table for any audit planning you are interested in. Find ways to understand auditors’ challenges, pain points and articulate what’s possible. More importantly, take the opportunity to speak with others about analytics, both formally and informally, fostering a more favorable data culture within the team.
Perfect is the Enemy of Good
Data scientists have a natural obsession with models and tools. In an audit setting, this has to be balanced with “speed to insights” and sustainability.
Firstly, audit projects usually run in months, a significant portion of which is spent on back-and-forth discussions between auditors and auditees. Insights derived from complex models are not intended to be perfect from the beginning, but they are expected to more effectively facilitate auditors’ deep dives into areas of audit interest. In these deep dives, feedback from auditees, who are closer to the business context of data, plays a greater role in transforming data insights into audit findings. Therefore, the timeliness of insights delivered to auditors matters more than the sophistication of models being used.
Secondly, in the hypes of ever-increasing new technologies, auditors tend to look to adoption of new tools as proof that data analytics are being done. Such tendency can be pushed top-down from audit leadership in the drive to be “digital” in every corner of the organization. But in an enterprise environment, internal audit, as a back-office function, do not possess the same level of technology resources and support as front-office departments. In the rush to embrace new technologies, auditors and accompanying resources might be exhausted before reaping any practical values from these technologies.
Tips for data scientists: 1) Work closely with auditors to align on project priorities and timeline, and prepare to be part of the discussions with auditees and iterate your deliverables as many times as needed to formulate the final audit opinions/findings. 2) Aim big but start small: break down your ambitious “digitalization” plans into a series of small, incremental steps rather than giant leaps forward, and where possible, plan for some “quick wins” (i.e., analytics initiatives for auditors to gain short-term initiatives).
Achieving Audit Objectives with Greater Impact
However fancy data science is pictured, it is primarily a tool to supercharge auditors to achieve audit objectives with greater impact. While we look for the best tool for the job, we should not focus too much on the tool and too little on the job.