How to use Google searches to spot corporate fraud

Each internet search leaves traces that tell a story about market demand, and a new study reveals its link to exposing fraud

Featured faculty: Zhang Yinglei
Written by Ellis Ng

Earnings reports are the scoreboard of a company’s financial health. However, behind those soaring revenue numbers, the truth can sometimes be buried. Some firms inflate their revenues to dazzle investors. It’s a deceptive game that brought giants like the payment company Wirecard and the property developer Evergrande to the ground.

Revenue misstatement can be difficult to spot. Firms often have complex operations and various ways of recording revenue, which can obscure the true financial picture. Traditional audit procedures may not always detect subtle manipulations, particularly when management crafts the presentation of revenues to fit with strategic narratives.

This challenge has led auditors to seek independent and external sources to help spot potential revenue misstatements. A new study suggests that Google can provide answers, given the search engine’s unique feature of collecting user queries.

“Customers nowadays seek product information, such as quality reviews, outlets, and pricing, before making a purchase. As the largest search engine globally, Google serves as a valuable resource to aggregate demand information,” says Zhang Yinglei, Associate Professor at the School of Accountancy at the Chinese University of Hong Kong (CUHK) Business School.

In a study titled Using Google searches of firm products to detect revenue management, Professor Zhang and her collaborators introduce the manipulation up (MUP) model by harnessing big data. The model leverages the search volume index provided by Google Trends that quantifies the relative frequency of searches for a particular term and compares it with the company’s reported sales figures.

“If reported sales growth significantly rises while Google search volume for the firm’s products significantly declines, this discrepancy should raise a red flag for auditors, analysts, and fraud investigators,” she says.

With this model, auditors and fraud investigators can verify whether the reported revenue aligns with actual consumer interests. It also presents a low-cost and timely tool for auditors that leverages publicly accessible data to independently validate reported revenues and identify fraud.

Turning a search engine into a fraud finder

In developing the model, Professor Zhang, along with Chiu Peng Chia of CUHK Shenzhen, Teoh Siew Hong of the University of California, Los Angeles, and Huang Xuan of California State University, first gathered quarterly firm-level sales data and other financial variables from S&P Compustat, which hosts a comprehensive market and corporate financial database.

They also collect Google Trends data to obtain the search volume index for each firm’s products from 2004 to 2020. The researchers then picked each firm’s biggest advertised brand, collected monthly search data for those products, and then compared changes in search interest to the companies’ reported sales to see if significant differences could signal possible revenue manipulation.

“If reported sales growth significantly rises while Google search volume for the firm’s products significantly declines, this discrepancy should raise a red flag for auditors, analysts, and fraud investigators.”

Professor Zhang Yinglei

The analyses confirm that changes in search volume index are positively correlated with actual changes in sales, validating the use of the MUP model as an external measurement correlating with firm revenue. What makes the tool particularly valuable is that Google Trends data can’t be manipulated by company management, making it an impartial source for genuine consumer interest.

“Google search data offers an independent way to verify company claims, rather than relying solely on information that companies can control,” says Professor Zhang.

The missing piece of the puzzle

To gauge the MUP model’s reliability, the researchers further compare its incremental predictability through rigorous testing against four established fraud detection methods in financial audit, such as the F_Score that spots financial anomalies, Stubben’s discretionary revenues model that tracks unusual revenue patterns, analyst sales forecasts, and media and analyst coverage.

The study shows that the MUP model adds unique insights on top of these traditional tools. It also matches with other common warnings of questionable revenue reporting, such as unusual increases in unpaid customer bills (accounts receivable) and suspiciously low funds to cover losses from unpaid bills (bad debt provisions).

Nevertheless, the MUP model is designed as a prototype diagnostic tool that integrates external big data with traditional fraud risk indicators, but it is not intended to be used independently or as a substitute for other fraud detection methods. Thus, this tool is best used as part of multiple approaches rather than relying on it exclusively.

“The MUP model enhances cross-checking methods by adding a source that is hard for companies to manipulate and helping auditors spot problems during quarterly reviews and in industries where customer search behaviour reflects demand strongly,” Professor Zhang adds.

Simply put, this model is most effective for cross-checking revenue reports in the retail and business-to-customer industries, where customers are more likely to search for product information before purchase. By combining data from search engines with traditional financial analysis, auditors can build a more complete picture of a company’s true performance.

The future of fraud detection

Auditors aren’t the only ones to benefit from the MUP model. Investors can also use it to evaluate company risks before making investment decisions, while regulators and policymakers can incorporate it into their fraud prevention strategies.

By analysing Google search data not just for a single product but also for companies within the same industry, analysts and auditors can gauge the relationship between the company’s revenue changes and its competitors. This could help make more accurate and improve the ability to spot fraud.

Comparing competitors’ Google Trends data will also enable analysts to distinguish between company-specific issues and industry-wide trends, helping determine whether declining consumer interest is unique to one company or reflects broader market changes.

What about markets where Google is less commonly used? Professor Zhang believes that the approach works with other search engines. “For example, fraud investigators can utilise the Baidu search index in Chinese Mainland and the Naver search index in South Korea to assess genuine consumer interest in a firm’s products.”

The study further advocates for regulators to adopt a more proactive stance in combating fraud by collecting transaction data from independent external sources and making this information publicly available. Such collaborative efforts would make fraud detection methods more accessible and affordable for auditors and stakeholders.

Easing the identification of potential financial misstatements would diminish the incentives for revenue manipulation, thereby discouraging such unethical behaviour. Ultimately, transparency stands as a powerful deterrent, perhaps the most effective safeguard, against fraudulent behaviours.