�鶹��ý

The days of accountants relying on sprawling legal notepads, bulky calculators, and illegible chicken scratch are long gone. Accounting in 2021, like many industries, has been revolutionized by an onslaught of new technologies�鶹��ý�ranging from complex Microsoft Excel spreadsheets and formulas to ever-sophisticated software.

Pace Accounting Professor Freddy Huang is well aware of these changes, and is helping to usher in a new era in accounting. Through the Multi-Dimensional Audit Data Selection (MADS) framework�鶹��ý�which Huang and his colleagues helped develop in conjunction with the American Institute of Certified Public Accountants (AICPA), the Chartered Professional Accountants (CPA) Canada, and audit experts from the Big Four accounting firms�鶹��ý�Huang is helping auditors to, quite literally, crunch the numbers more effectively.

A large component of accounting is auditing; which refers to independent examination of financial information of a company, often conducted by external accounting firm. For large companies, auditing theoretically involves looking at potentially millions of business transactions in a given year.

�鶹��ý�Traditionally, auditing is quite labor intensive,�鶹��ý� said Huang.

Because of this reality, auditors have long engaged in a process known as sampling�鶹��ý�which involves taking a smaller portion of transactions and analyzing that portion to make determinations about the full quantity of transactions.

�鶹��ý�In relation to audit data analytics, it�鶹��ý�s important for students to develop a mindset."

�鶹��ý�They don�鶹��ý�t take look at all of a business�鶹��ý� transactions. They�鶹��ý�ll just select, let�鶹��ý�s say, 100 records as a representative sample of the population,�鶹��ý� said Huang. �鶹��ý�Then, with those records, they�鶹��ý�ll apply some audit procedures, and draw conclusions on the entire population.�鶹��ý�

However, as Huang notes, this method has historically come with a risk, which is commonly known as�鶹��ý�you guessed it�鶹��ý�sampling risk.

�鶹��ý�If you only audit 100 records, your conclusion may be different from if you hypothetically audit the full population (thousands or possibly millions of records).�鶹��ý�

Enter, data analytics. With the advantage of today�鶹��ý�s data analytic tools, it has become increasingly easier and cost-effective to examine large quantities of transactions with a relatively low-cost, which theoretically solves the problems of sampling risk. Yet, a new problem has emerged. Namely, if auditors have several weeks to audit a company, what is the most effective way to break down this deluge of information? Is even attempting to sift through mountains of data in an auditor�鶹��ý�s best interest?

�鶹��ý�One reason why accounting firms are hesitant to apply full population tests are the large number of outliers,�鶹��ý� notes Huang.

Outliers, are in essence, the red flags�鶹��ý�transactions in an audit that based on the audit criteria, raise possible suspicion and merit a further look. Because a full-scale audit can produce thousands of outliers, analyzing all of the outliers in a large audit is often impractical, and possibly conducive the human error.

The key is then, as Huang explains, to apply frameworks to further reduce the number of outliers down a number that is manageable for an auditor, while also ensuring those outliers are the most important transactions for the auditor to look at. This is exactly the problem that Huang�鶹��ý�s Multi-Dimensional Audit Data Selection (MADS) Framework is dedicated to solving, and does so by breaking down outliers into three different outputs. In other words, it uses algorithms and equations to even further reduce the number of outliers in an audit to the ones that raise the greatest suspicion.

�鶹��ý�We first start with the full population analysis,�鶹��ý� explains Huang. �鶹��ý�Based on whatever test or filters you need to apply, you�鶹��ý�ll first get a large set of outliers. That�鶹��ý�s the Stage One output, but not the final output of the analysis. We then apply additional tests for our Stage Two output, which could be an advanced machine learning algorithm, or based on the output of Stage One, auditors may develop additional filters or test using their own judgement.�鶹��ý�

Finally, in the Stage Three output, Huang and his colleagues created a �鶹��ý�suspicion score,�鶹��ý� which evolves an equation that weights transactions based on the feature and/or number of violations a given transaction has along with the dollar amount of a transaction. If two transactions have the same amount and type of violations but one transaction is significantly larger in dollar amount, for example, that transaction would most likely be greater cause for concern.

Ultimately, through the MADS framework�鶹��ý�as well as through looking at how to best incorporate Robotic Process Automation (RPA) into auditing (which aims to free up accountants from doing repetitive and low-judgement tasks in order to focus on tasks that require greater professional judgment)�鶹��ý�Huang understands that the future of accounting is heavily trending toward having a strong command of data analytics. When it comes to training the next generation of accountants at Pace, Huang believes that ensuring students approach complex tasks with a strong command of today�鶹��ý�s technological tools is paramount for success.

�鶹��ý�In relation to audit data analytics, it�鶹��ý�s important for students to develop a mindset. In the future, when they start to work for different accounting firms and with different clients, they will certainly be dealing with different types of data sets. But with this mindset, they'll always know where to start and what to follow."