Fraud Detection in Financial Statements: How Algorithms Complement Controls

The annual financial statements mark the end of an intensive period of work. The figures must be correct. The reports must be accurate. Traditionally, fraud detection in the financial statements of an SME relies on established processes and the four-eyes principle. These manual controls are the foundation of the internal control system. However, they reach their limits when it comes to analysing large volumes of data and interpreting texts. New methodological approaches can offer a valuable addition here to strengthen human review.

The Limits of Manual Control

The internal control system (ICS) and the four-eyes principle are indispensable. They ensure the quality of financial reporting and build trust. Experienced controllers and trustees often recognise discrepancies through their many years of practice. This human expertise cannot be completely replaced by any machine.

However, manual review has natural limitations. The human eye can tire. Cognitive biases can unknowingly influence perception. These include confirmation bias, where one primarily seeks information that supports an existing assumption, or the anchoring effect, which leads to an excessive weighting of the first piece of information received. With tens of thousands of journal entry lines and complex reports, it is practically impossible to check every detail with the same level of attention. Subtle but systematic deviations may thus go undetected.

The complexity of the business world is also constantly increasing. ERP systems generate growing amounts of data. At the same time, the narrative parts of annual reports, such as the management report, are becoming more extensive. This development presents new challenges for traditional control mechanisms, as the volume often exceeds the capacity for a manual in-depth review.

What the Language in Financial Reports Reveals

The way management communicates is itself a source of information. A new study has systematically investigated this connection (Journal of Management and Business Solutions 2024). The researchers analysed the tone of management in financial reports as well as the complexity of the language used.

The results are insightful. The study used Natural Language Processing methods to measure tone. It was found that both an overly positive tone and a strikingly negative tone correlated significantly with fraudulent financial reporting (Journal of Management and Business Solutions 2024). An excessively optimistic tone can be an attempt to distract from poor results. An unusually negative tone may indicate an attempt to lower expectations sharply to make future results look better.

Another important factor was the complexity of the reporting. Complicated and difficult-to-understand language can be deliberately used to obscure facts and make a detailed audit more difficult. The study confirmed that the complexity of financial reports is a strong predictor of fraudulent intent (Journal of Management and Business Solutions 2024). Clarity and transparency are therefore not just a matter of style, but also a feature of good corporate governance.

Algorithms as a Supplement, Not a Replacement

These technological approaches do not replace human judgement. They serve as a powerful tool to supplement the expertise of finance professionals. Algorithms can objectively and tirelessly search large volumes of data and text for predefined patterns. They identify outliers that might escape manual review and direct the attention of reviewers to the areas of highest risk.

In the aforementioned study, various anomaly detection algorithms were evaluated. A so-called Autoencoder Neural Network showed high predictive performance. It achieved an accuracy of 93 per cent and an AUC value of 0.96 in identifying suspicious financial statements (Journal of Management and Business Solutions 2024). The AUC value measures the model's ability to distinguish between positive and negative cases and confirms its high reliability.

These machine learning-based techniques were significantly superior to traditional statistical models such as logistic regression (Journal of Management and Business Solutions 2024). The reason is that they are better able to recognise complex and non-linear relationships in the data. Fraudulent patterns are rarely simple and linear, but are often hidden in multi-layered data structures.

The researchers conclude from their results that the integration of management tone analysis, report complexity indicators, and anomaly detection algorithms significantly improves the detection of fraud in financial statements (Journal of Management and Business Solutions 2024). Human control is strengthened by a layer of machine analysis.

Stability in Uncertain Times

Strengthening internal controls is particularly relevant in an uncertain economic environment. Such phases can increase the pressure on management to achieve set targets, which can increase the risk of opportunistic behaviour. The digital transformation can make a significant contribution to risk mitigation here. A study shows that digitalisation improves the quality of accounting information positively and statistically significantly (Accounting, Finance and Computational Intelligence 2024).

Digital technologies increase transparency, reduce information asymmetries and restrict opportunistic management behaviour (Accounting, Finance and Computational Intelligence 2024). This effect is particularly pronounced under unstable business conditions, as robust digital processes are less susceptible to ad-hoc adjustments. In uncertain times, the reliability of financial information is all the more important for the trust of investors, lenders and other stakeholders.

Solid reporting also has macroeconomic effects. Conservative accounting, which reports profits rather cautiously, can contribute to macroeconomic stability. A study found that such prudent reporting has a negative and statistically significant effect on economic volatility (Accounting, Finance and Computational Intelligence 2023). Reliable and transparent financial statements are therefore not only in the interest of the individual company, but also strengthen confidence in the economy as a whole.

Practical Steps for Your Internal Control System

The implementation of such analyses does not have to be complex. There are concrete steps you can take today within your finance team to increase the effectiveness of your controls.

• Start by taking stock of your existing processes. Document the chain from the ERP export to the final report. Where do controls rely exclusively on manual sampling? At what points does the volume of data make a complete review difficult?
• Analyse the narrative parts of your reports from recent years. Take the management report. Has the language remained consistent over the years or has it become more complex? Has the tone changed? Such a qualitative assessment raises awareness of the soft factors.
• Use the tools you already have. Standard software like Microsoft Excel offers powerful data preparation and analysis functions with tools like Power Query, which often go unused. Conditional formatting, pivot tables or sorting functions can help to identify initial anomalies in large data sets.
• Define clear criteria for anomalies in your company. An anomaly could be a journal entry above a certain amount, a significant deviation from a budget, or a transaction at an unusual time. Without a clear definition, even the best technology cannot deliver meaningful results.
• Foster a culture of professional scepticism. Encourage all team members to ask questions and investigate discrepancies, regardless of their size. Technology is a tool, but the critical attitude of finance professionals remains the decisive factor.

Microsoft and Excel are registered trademarks of the Microsoft Corporation.